Previous value: -"Optional already-uploaded research files. Accepts a public HTTP(S) URL or a file already in the user's Minds uploads under chat/<userId>/ (its storage path or its /api/uploads/chat/ URL). MCP cannot read a local path. Files shared in the conversation (uploaded or generated images, PDFs, documents) go in the tool's top-level conversationFiles field where the tool has one; elsewhere a PDF or image without a public URL has no MCP route and is uploaded in the Minds app first. Plain text contents (UTF-8 text, Markdown, CSV, JSON) can go to import_audience_sources. External URLs are copied into Minds storage before a Study is saved or run. temp/, portfolio/ and context/ paths and /api/uploads/file-access/ URLs belong to other owners and are not accepted. A Study does not start until Minds can read every file. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."New value: +"Optional already-uploaded research files. Accepts a public HTTP(S) URL or a file already in the user's Minds uploads under chat/<userId>/ (its storage path or its /api/uploads/chat/ URL). MCP cannot read a local path. Files shared in the conversation (uploaded or generated images, PDFs, documents) go in the tool's top-level conversationFiles field where the tool has one; elsewhere a PDF or image without a public URL has no MCP route and is uploaded in the Minds app first. Plain text contents (UTF-8 text, Markdown, CSV, JSON) can go to import_audience_sources. External URLs are copied into Minds storage before a Study is saved or run. temp/, portfolio/ and context/ paths and /api/uploads/file-access/ URLs belong to other owners and are not accepted. A Study does not start until Minds can read every file. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance. Every file shapes the grounding, but only one classified as supporting material (not a spreadsheet) enters the members' own knowledge; the result's withheldFromMemberKnowledge names each other file with its reason (respondent_dataset, study_report, questionnaire, screener, distribution_source, tabular, or unclassified: not classified as supporting material, also when the reading failed or ran out of time)."

Previous value: -"How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Pro or Enterprise plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account."New value: +"How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It builds at most 20 Minds; a larger memberCount moves to a deeper mode your plan includes, else is built at 20 with a note. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two are included only on some plans (get_audience_limits shows which apply to you) — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account."

Previous value: -"Exact number of Minds to create. Pass this whenever the user states a size (\"exactly 50 per region\", \"genau 50 Minds je Zelle\") instead of relying on the server to parse the number out of the brief. When omitted, the size is inferred from the brief and, failing that, from the creation mode's evidence-based automatic sizing. Two separate refusals apply: 403 MODE_CAP above 20 in \"balanced\" mode (switch to segment_coverage or benchmark_depth), and 403 PLAN_LIMIT above your plan's per-Audience cap. The 6000 accepted here is the absolute system ceiling, not your allowance — call get_audience_limits for the real one. The Audience is never created at a partial size."New value: +"Exact number of Minds to create. Pass this whenever the user states a size (\"exactly 50 per region\", \"genau 50 Minds je Zelle\") instead of relying on the server to parse the number out of the brief. When omitted, the size is inferred from the brief and, failing that, from the creation mode's evidence-based automatic sizing. The only refusal is 403 PLAN_LIMIT above your plan's per-Audience cap; above 20 in \"balanced\" see audienceCreationMode. The 6000 accepted here is the absolute system ceiling, not your allowance — call get_audience_limits for the real one. Apart from that note, the Audience is never created at a partial size."

Previous value: -"Set false to skip Exa web search and extraction completely. The Audience is then grounded only on the brief and supplied files, links, keywords, and structurally parsed respondent data. Note that a study report, questionnaire, screener or respondent dataset grounds the Audience but is deliberately withheld from the members' own knowledge, so it cannot be the evidence they answer from; with nothing left to ground on, creation is refused rather than producing Minds with no evidence. Leave it true unless the supplied sources include material the members themselves should read."New value: +"Set false to skip Exa web search and extraction completely. The Audience is then grounded only on the brief and supplied files, links, keywords, and structurally parsed respondent data. Note that a study report, questionnaire, screener or respondent dataset grounds the Audience but is deliberately withheld from the members' own knowledge, so it cannot be the evidence they answer from; with nothing else to ground on, the Audience is still created from the brief and its charts and says that its members carry no readable evidence. Leave it true unless the supplied sources include material the members themselves should read."

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs, including respondent datasets, default to the full webapp draft with proposed Minds; only a legacy full dataset payload uses a quota-only review. Setting false without reviewed evidence still returns a preview. Present the plan and wait until the user confirms it, then set grounding.preview:false with the unchanged reviewedGroundingJson/SHA pair and the same allocation options. Supplying saved evidence without an explicit false keeps it in review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs, including respondent datasets, default to the full webapp draft with proposed Minds; only a legacy full dataset payload uses a quota-only review. Setting false without reviewed evidence still returns a preview. Present the plan and wait until the user confirms it, then set grounding.preview:false with the unchanged reviewedGroundingJson/SHA pair and the same allocation options. Supplying saved evidence without an explicit false keeps it in review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes are set aside and listed in cohortAllocation.audit.warnings (or as a target_unapplied notice on the grounding), never refused. If research did not finish, the preview says so (research_unavailable) and creating from it researches the Audience in the background."

Previous value: -"When a completed operation is read back: member rollup from GET /api/v1/audiences/{id}/progress. isReady = every Mind can chat (the Audience is ready); learning = ready Minds still learning from their sources in the background; fullyTrained = none still learning; research.phase = researching while the Audience's background research runs (it is usable meanwhile), researched once merged and applied."New value: +"When a completed operation is read back: member rollup from GET /api/v1/audiences/{id}/progress. isReady = every Mind can chat (the Audience is ready); startsNow = a Study started now runs at once, without leftOut Minds whose training failed for good; learning = ready Minds still learning from their sources in the background; fullyTrained = none still learning; research.phase = researching while the Audience's background research runs (it is usable meanwhile), researched once merged and applied."

Added value: +{
+  "anyOf": [
+    {
+      "items": {},
+      "type": "array"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "description": "The caller's existing Audiences with exactly this name (id, name, workspaceUrl), when a new brief was started; this call still creates a separate Audience."
+}

Previous value: -"Reviewed variable roles, distributions, and privacy-safe pairwise relationships returned by preview_audience_dataset_segmentation. Requires audienceCreationMode=\"benchmark_depth\". Structural variables shape one generalizable representative cohort; outcomes remain held out and joint combinations remain audit evidence only. Copy `respondentCount` and `recommendedMindCount` through from the preview — they cap the cohort size."New value: +"Which variables of a respondent dataset shape the cohort. A draft whose research.files include a respondent dataset (XLS, XLSX, CSV, SPSS .sav; Pro plan, audienceCreationMode=\"benchmark_depth\") reads it and proposes this selection itself; its review lists every variable with its role and observed shares. Omit this on the first draft. To change the variables, revise that draft (draft.preview:true with its id and sha256, the same brief and research.files) passing {fileName, sheetName, segmentationColumns} from the review with the keys the user chose. Structural variables shape one representative cohort; outcomes stay held out."

Previous value: -"Column-level distributions from the preview. Used when `variables` is absent."New value: +"Legacy: column-level distributions, used when `variables` is absent. Omit with a selection."

Previous value: -"Observed joint profiles as audit counts. Copy the ids and counts from the preview; per-cell values are not needed."New value: +"Legacy: observed joint profiles as audit counts. Omit with a selection."

Added value: +"The dataset's file name, as the draft's datasetSegmentation review names it. The same file must be in research.files (or conversationFiles)."

Previous value: -"Preview-recommended cohort size. Caps how many Minds the benchmark run creates."New value: +"Legacy: preview-recommended cohort size, which caps the cohort. Omit with a selection."

Previous value: -"Total respondent rows in the reviewed sheet, copied from the preview."New value: +"Legacy: total respondent rows in the reviewed sheet. Omit with a selection."

Previous value: -"Variable-level distributions from the preview (preferred over `columns` when present)."New value: +"Legacy: variable-level distributions (preferred over `columns` when present). Omit with a selection."

Previous value: -[
-  "fileName",
-  "segmentationColumns",
-  "combinations"
-]New value: +[
+  "fileName",
+  "segmentationColumns"
+]

Previous value: -"Deprecated; omit it. Every member of every Audience is trained individually in the background, including reviewed-dataset cohorts of more than ~40 Minds (each keeps its deterministic cohort profile and the request locale). true is accepted as a no-op; false is rejected, because it used to create untrained Minds with an empty system prompt that reported ready. Cost and time: one Mind training per member; a 100-Mind cohort takes tens of minutes to a few hours. The tool returns as soon as the Audience exists (structuredContent.audience.memberTraining reports what was queued); wait until every member reports readyToChat (get_mind_status, GET /api/v1/minds/{id}/training, or readiness.isReady from GET /api/v1/audiences/{id}/progress) before running a Study, which refuses Audiences with untrained members; the Audience is ready at that point. Members keep learning from their sources in the background for a few more minutes (readiness.learning, then readiness.fullyTrained); nothing needs to wait for it. readyToChat means the member can answer, not that it researched anything: a member whose collection found nothing keeps one generated context item and still reports ready, so check knowledgeState.fallbackOnly on the members before treating the Audience as evidence-backed."New value: +"Deprecated; omit it. Every member of every Audience is trained individually in the background, including reviewed-dataset cohorts of more than ~40 Minds (each keeps its deterministic cohort profile and the request locale). true is accepted as a no-op; false is rejected, because it used to create untrained Minds with an empty system prompt that reported ready. Cost and time: one Mind training per member; a 100-Mind cohort takes tens of minutes to a few hours. The tool returns as soon as the Audience exists (structuredContent.audience.memberTraining reports what was queued); wait until every member reports readyToChat (readiness.isReady when this tool is called again with the returned operationId) before running a Study, which refuses Audiences with untrained members; the Audience is ready at that point. Members keep learning from their sources in the background for a few more minutes (readiness.learning, then readiness.fullyTrained); nothing needs to wait for it. readyToChat means the member can answer, not that it researched anything: a member whose collection found nothing keeps one generated context item and still reports ready, so check knowledgeState.fallbackOnly on the members before treating the Audience as evidence-backed."

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs default to the full webapp draft with proposed Minds; reviewed respondent datasets use a quota-only review. Setting false without reviewed evidence still returns a preview. Present the plan and wait until the user confirms it, then set grounding.preview:false with the unchanged reviewedGroundingJson/SHA pair and the same allocation options. Supplying saved evidence without an explicit false keeps it in review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs, including respondent datasets, default to the full webapp draft with proposed Minds; only a legacy full dataset payload uses a quota-only review. Setting false without reviewed evidence still returns a preview. Present the plan and wait until the user confirms it, then set grounding.preview:false with the unchanged reviewedGroundingJson/SHA pair and the same allocation options. Supplying saved evidence without an explicit false keeps it in review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed."

Added value: +{
+  "description": "Deprecated; omit it. Every member of every Audience is trained individually in the background, including reviewed-dataset cohorts of more than ~40 Minds (each keeps its deterministic cohort profile and the request locale). true is accepted as a no-op; false is rejected, because it used to create untrained Minds with an empty system prompt that reported ready. Cost and time: one Mind training per member; a 100-Mind cohort takes tens of minutes to a few hours. The tool returns as soon as the Audience exists (structuredContent.audience.memberTraining reports what was queued); wait until every member reports readyToChat (get_mind_status, GET /api/v1/minds/{id}/training, or readiness.isReady from GET /api/v1/audiences/{id}/progress) before running a Study, which refuses Audiences with untrained members; the Audience is ready at that point. Members keep learning from their sources in the background for a few more minutes (readiness.learning, then readiness.fullyTrained); nothing needs to wait for it. readyToChat means the member can answer, not that it researched anything: a member whose collection found nothing keeps one generated context item and still reports ready, so check knowledgeState.fallbackOnly on the members before treating the Audience as evidence-backed.",
+  "type": "boolean"
+}

Previous value: -"The webapp Audience draft workflow. Preview proposes Minds before Create; id and sha256 identify the exact reviewed roster. Rework revises that draft. Set preview:false with both id and sha256 only after the user confirms the displayed draft."New value: +"The webapp Audience draft workflow. Preview proposes Minds before Create; id and sha256 identify the exact reviewed roster. Rework revises that draft. Set preview:false with both id and sha256 only after the user confirms the displayed draft; that confirmation needs nothing else (omit the brief and every other option): the server creates exactly the reviewed draft from its stored request."

Previous value: -"How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account."New value: +"How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Pro or Enterprise plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account."

Previous value: -"The webapp Audience draft workflow. Preview proposes real Minds before Create; id and sha256 identify the exact reviewed roster. Rework revises that draft."New value: +"The webapp Audience draft workflow. Preview proposes Minds before Create; id and sha256 identify the exact reviewed roster. Rework revises that draft. Set preview:false with both id and sha256 only after the user confirms the displayed draft."

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs default to the full webapp draft with proposed Minds; set grounding.preview:true only for a quota-only review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs default to the full webapp draft with proposed Minds; reviewed respondent datasets use a quota-only review. Setting false without reviewed evidence still returns a preview. Present the plan and wait until the user confirms it, then set grounding.preview:false with the unchanged reviewedGroundingJson/SHA pair and the same allocation options. Supplying saved evidence without an explicit false keeps it in review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed."

Removed value: -[
-  {
-    "type": "string"
-  },
-  {
-    "type": "null"
-  }
-]

Added value: +[
+  "string",
+  "null"
+]

Added value: +{
+  "description": "Research files (screener, questionnaire, report, respondent data), read like research.files. Files shared in this conversation: uploaded or generated images, PDFs, documents. Each is copied into Minds storage on receipt; its file_id is its attachment id.",
+  "items": {
+    "properties": {
+      "download_url": {
+        "description": "Download URL the chat host supplies for this file.",
+        "minLength": 1,
+        "type": "string"
+      },
+      "file_id": {
+        "description": "The chat host's ID for this file.",
+        "minLength": 1,
+        "type": "string"
+      },
+      "file_name": {
+        "description": "File name when the host supplies it.",
+        "minLength": 1,
+        "type": "string"
+      },
+      "mime_type": {
+        "description": "MIME type when the host supplies it.",
+        "minLength": 1,
+        "type": "string"
+      }
+    },
+    "required": [
+      "download_url",
+      "file_id"
+    ],
+    "type": "object"
+  },
+  "type": "array"
+}

Previous value: -"Optional already-uploaded research files. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path. A file shared in the conversation (uploaded or generated image, PDF, document) is read through the tool's top-level files field where the tool has one; elsewhere a PDF or image with no public URL has no MCP route and is uploaded in the Minds app first. Text contents (UTF-8 text, Markdown, CSV, JSON) can be sent to import_audience_sources. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path, and any /api/uploads/file-access/ URL, is refused as not belonging to the workspace owner; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."New value: +"Optional already-uploaded research files. Accepts a public HTTP(S) URL or a file already in the user's Minds uploads under chat/<userId>/ (its storage path or its /api/uploads/chat/ URL). MCP cannot read a local path. Files shared in the conversation (uploaded or generated images, PDFs, documents) go in the tool's top-level conversationFiles field where the tool has one; elsewhere a PDF or image without a public URL has no MCP route and is uploaded in the Minds app first. Plain text contents (UTF-8 text, Markdown, CSV, JSON) can go to import_audience_sources. External URLs are copied into Minds storage before a Study is saved or run. temp/, portfolio/ and context/ paths and /api/uploads/file-access/ URLs belong to other owners and are not accepted. A Study does not start until Minds can read every file. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."

Previous value: -"URL of an already-uploaded file. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path. A file shared in the conversation (uploaded or generated image, PDF, document) is read through the tool's top-level files field where the tool has one; elsewhere a PDF or image with no public URL has no MCP route and is uploaded in the Minds app first. Text contents (UTF-8 text, Markdown, CSV, JSON) can be sent to import_audience_sources. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path, and any /api/uploads/file-access/ URL, is refused as not belonging to the workspace owner; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset."New value: +"URL of an already-uploaded file. Accepts a public HTTP(S) URL or a file already in the user's Minds uploads under chat/<userId>/ (its storage path or its /api/uploads/chat/ URL). MCP cannot read a local path. Files shared in the conversation (uploaded or generated images, PDFs, documents) go in the tool's top-level conversationFiles field where the tool has one; elsewhere a PDF or image without a public URL has no MCP route and is uploaded in the Minds app first. Plain text contents (UTF-8 text, Markdown, CSV, JSON) can go to import_audience_sources. External URLs are copied into Minds storage before a Study is saved or run. temp/, portfolio/ and context/ paths and /api/uploads/file-access/ URLs belong to other owners and are not accepted. A Study does not start until Minds can read every file."

Previous value: -"Optional already-uploaded research files. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path, and a file the user attached in the chat client is NOT reachable either unless that client also exposes a public URL for it — most do not. When it is not: send the file CONTENTS to import_audience_sources (UTF-8 text, Markdown, CSV, JSON), which takes contents rather than paths. A PDF or image with no public URL has no MCP route at all — it must be uploaded in the Minds app first; say so instead of retrying. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."New value: +"Optional already-uploaded research files. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path. A file shared in the conversation (uploaded or generated image, PDF, document) is read through the tool's top-level files field where the tool has one; elsewhere a PDF or image with no public URL has no MCP route and is uploaded in the Minds app first. Text contents (UTF-8 text, Markdown, CSV, JSON) can be sent to import_audience_sources. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path, and any /api/uploads/file-access/ URL, is refused as not belonging to the workspace owner; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."

Previous value: -"URL of an already-uploaded file. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path, and a file the user attached in the chat client is NOT reachable either unless that client also exposes a public URL for it — most do not. When it is not: send the file CONTENTS to import_audience_sources (UTF-8 text, Markdown, CSV, JSON), which takes contents rather than paths. A PDF or image with no public URL has no MCP route at all — it must be uploaded in the Minds app first; say so instead of retrying. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset."New value: +"URL of an already-uploaded file. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path. A file shared in the conversation (uploaded or generated image, PDF, document) is read through the tool's top-level files field where the tool has one; elsewhere a PDF or image with no public URL has no MCP route and is uploaded in the Minds app first. Text contents (UTF-8 text, Markdown, CSV, JSON) can be sent to import_audience_sources. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path, and any /api/uploads/file-access/ URL, is refused as not belonging to the workspace owner; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset."

Added value: +{
+  "description": "The webapp Audience draft workflow. Preview proposes real Minds before Create; id and sha256 identify the exact reviewed roster. Rework revises that draft.",
+  "properties": {
+    "id": {
+      "format": "uuid",
+      "pattern": "^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$",
+      "type": "string"
+    },
+    "preview": {
+      "type": "boolean"
+    },
+    "rework": {
+      "maxLength": 200000,
+      "type": "string"
+    },
+    "sha256": {
+      "pattern": "^[a-f0-9]{64}$",
+      "type": "string"
+    }
+  },
+  "type": "object"
+}

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool polls it to completion. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool returns its operation immediately so the widget can stream sources and distributions. New briefs default to the full webapp draft with proposed Minds; set grounding.preview:true only for a quota-only review. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."

Previous value: -"Resume a queued preview or creation by its returned jobId. Reads its v1 operation status without creating or charging anything. Supply this alone; do not repeat the creation brief to check progress."New value: +"Resume a queued preview or creation by its returned jobId. Reads its v1 operation status without creating or charging anything. Supply this alone; do not repeat the creation brief to check progress. Once the creation completed, the result also says whether every Mind can chat yet (the Audience is ready then) and how many are still learning from their sources in the background; call it again to follow that."

Added value: +{
+  "description": "Reviewed Audience brief, distributions and creation options for the widget."
+}

Added value: +{
+  "description": "Research and allocation preview; no Audience or Minds have been created."
+}

Added value: +{
+  "description": "When a completed operation is read back: member rollup from GET /api/v1/audiences/{id}/progress. isReady = every Mind can chat (the Audience is ready); learning = ready Minds still learning from their sources in the background; fullyTrained = none still learning; research.phase = researching while the Audience's background research runs (it is usable meanwhile), researched once merged and applied."
+}

Previous value: -"Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."New value: +"Optional already-uploaded research files. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path, and a file the user attached in the chat client is NOT reachable either unless that client also exposes a public URL for it — most do not. When it is not: send the file CONTENTS to import_audience_sources (UTF-8 text, Markdown, CSV, JSON), which takes contents rather than paths. A PDF or image with no public URL has no MCP route at all — it must be uploaded in the Minds app first; say so instead of retrying. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."

Previous value: -"URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset."New value: +"URL of an already-uploaded file. Pass a fetchable HTTP(S) URL, or an existing Minds workspace upload under chat/<userId>/ (that storage path or its /api/uploads/chat/ URL; request the signed upload with folder \"chat\"). MCP cannot read a local path, and a file the user attached in the chat client is NOT reachable either unless that client also exposes a public URL for it — most do not. When it is not: send the file CONTENTS to import_audience_sources (UTF-8 text, Markdown, CSV, JSON), which takes contents rather than paths. A PDF or image with no public URL has no MCP route at all — it must be uploaded in the Minds app first; say so instead of retrying. Study tools import external file URLs into durable Minds storage before saving or running. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset."

Previous value: -"Set false to skip Exa web search and extraction completely. The Audience is then grounded only on the brief and supplied files, links, keywords, and structurally parsed respondent data."New value: +"Set false to skip Exa web search and extraction completely. The Audience is then grounded only on the brief and supplied files, links, keywords, and structurally parsed respondent data. Note that a study report, questionnaire, screener or respondent dataset grounds the Audience but is deliberately withheld from the members' own knowledge, so it cannot be the evidence they answer from; with nothing left to ground on, creation is refused rather than producing Minds with no evidence. Leave it true unless the supplied sources include material the members themselves should read."

Removed value: -{
-  "description": "How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account.",
-  "enum": [
-    "balanced",
-    "segment_coverage",
-    "benchmark_depth"
-  ],
-  "type": "string"
-}

Removed value: -{
-  "description": "Deterministic allocation controls. Reviewed respondent datasets default to observed, which preserves the strongest privacy-safe structural relationships while keeping exact marginals; other sources default to independence. Use distributionNames/maxDistributions to choose axes, minSegmentCount for a feasible floor, and seed for reproducible external runs.",
-  "properties": {
-    "distributionNames": {
-      "items": {
-        "type": "string"
-      },
-      "maxItems": 500,
-      "type": "array"
-    },
-    "enabled": {
-      "type": "boolean"
-    },
-    "includeProfilesInResponse": {
-      "type": "boolean"
-    },
-    "jointStrategy": {
-      "enum": [
-        "independent",
-        "aligned",
-        "observed"
-      ],
-      "type": "string"
-    },
-    "maxDistributions": {
-      "maximum": 500,
-      "minimum": 1,
-      "type": "integer"
-    },
-    "minSegmentCount": {
-      "maximum": 20,
-      "minimum": 0,
-      "type": "integer"
-    },
-    "seed": {
-      "maxLength": 200,
-      "type": "string"
-    },
-    "targetOverrides": {
-      "items": {
-        "properties": {
-          "distribution": {
-            "type": "string"
-          },
-          "segments": {
-            "items": {
-              "properties": {
-                "label": {
-                  "type": "string"
-                },
-                "pct": {
-                  "maximum": 100,
-                  "minimum": 0,
-                  "type": "number"
-                }
-              },
-              "required": [
-                "label",
-                "pct"
-              ],
-              "type": "object"
-            },
-            "maxItems": 200,
-            "type": "array"
-          }
-        },
-        "required": [
-          "distribution",
-          "segments"
-        ],
-        "type": "object"
-      },
-      "maxItems": 500,
-      "type": "array"
-    }
-  },
-  "type": "object"
-}

Added value: +{
+  "description": "How many Minds to create and how the cohort is allocated across the grounded axes.",
+  "properties": {
+    "audienceCreationMode": {
+      "description": "How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account.",
+      "enum": [
+        "balanced",
+        "segment_coverage",
+        "benchmark_depth"
+      ],
+      "type": "string"
+    },
+    "cohortAllocation": {
+      "description": "Deterministic allocation controls. Reviewed respondent datasets default to observed, which preserves the strongest privacy-safe structural relationships while keeping exact marginals; other sources default to independence. Use distributionNames/maxDistributions to choose axes, minSegmentCount for a feasible floor, and seed for reproducible external runs.",
+      "properties": {
+        "distributionNames": {
+          "items": {
+            "type": "string"
+          },
+          "maxItems": 500,
+          "type": "array"
+        },
+        "enabled": {
+          "type": "boolean"
+        },
+        "includeProfilesInResponse": {
+          "type": "boolean"
+        },
+        "jointStrategy": {
+          "enum": [
+            "independent",
+            "aligned",
+            "observed"
+          ],
+          "type": "string"
+        },
+        "maxDistributions": {
+          "maximum": 500,
+          "minimum": 1,
+          "type": "integer"
+        },
+        "minSegmentCount": {
+          "maximum": 20,
+          "minimum": 0,
+          "type": "integer"
+        },
+        "seed": {
+          "maxLength": 200,
+          "type": "string"
+        },
+        "targetOverrides": {
+          "items": {
+            "properties": {
+              "distribution": {
+                "type": "string"
+              },
+              "segments": {
+                "items": {
+                  "properties": {
+                    "label": {
+                      "type": "string"
+                    },
+                    "pct": {
+                      "maximum": 100,
+                      "minimum": 0,
+                      "type": "number"
+                    }
+                  },
+                  "required": [
+                    "label",
+                    "pct"
+                  ],
+                  "type": "object"
+                },
+                "maxItems": 200,
+                "type": "array"
+              }
+            },
+            "required": [
+              "distribution",
+              "segments"
+            ],
+            "type": "object"
+          },
+          "maxItems": 500,
+          "type": "array"
+        }
+      },
+      "type": "object"
+    },
+    "datasetSegmentation": {
+      "description": "Reviewed variable roles, distributions, and privacy-safe pairwise relationships returned by preview_audience_dataset_segmentation. Requires audienceCreationMode=\"benchmark_depth\". Structural variables shape one generalizable representative cohort; outcomes remain held out and joint combinations remain audit evidence only. Copy `respondentCount` and `recommendedMindCount` through from the preview — they cap the cohort size.",
+      "properties": {
+        "columns": {
+          "description": "Column-level distributions from the preview. Used when `variables` is absent.",
+          "items": {
+            "properties": {
+              "distribution": {
+                "items": {
+                  "properties": {
+                    "respondentCount": {
+                      "maximum": 9007199254740991,
+                      "minimum": 0,
+                      "type": "integer"
+                    },
+                    "sharePct": {
+                      "minimum": 0,
+                      "type": "number"
+                    },
+                    "value": {
+                      "default": "",
+                      "type": "string"
+                    }
+                  },
+                  "required": [
+                    "value",
+                    "respondentCount",
+                    "sharePct"
+                  ],
+                  "type": "object"
+                },
+                "maxItems": 30,
+                "type": "array"
+              },
+              "key": {
+                "type": "string"
+              },
+              "label": {
+                "type": "string"
+              },
+              "otherRespondentCount": {
+                "maximum": 9007199254740991,
+                "minimum": 0,
+                "type": "integer"
+              },
+              "validRespondentCount": {
+                "maximum": 9007199254740991,
+                "minimum": 0,
+                "type": "integer"
+              }
+            },
+            "required": [
+              "key",
+              "label",
+              "distribution"
+            ],
+            "type": "object"
+          },
+          "maxItems": 200,
+          "type": "array"
+        },
+        "combinations": {
+          "description": "Observed joint profiles as audit counts. Copy the ids and counts from the preview; per-cell values are not needed.",
+          "items": {
+            "properties": {
+              "id": {
+                "type": "string"
+              },
+              "respondentCount": {
+                "exclusiveMinimum": 0,
+                "maximum": 9007199254740991,
+                "type": "integer"
+              },
+              "sharePct": {
+                "minimum": 0,
+                "type": "number"
+              }
+            },
+            "required": [
+              "id",
+              "respondentCount",
+              "sharePct"
+            ],
+            "type": "object"
+          },
+          "maxItems": 5000,
+          "type": "array"
+        },
+        "fileName": {
+          "type": "string"
+        },
+        "recommendedMindCount": {
+          "description": "Preview-recommended cohort size. Caps how many Minds the benchmark run creates.",
+          "exclusiveMinimum": 0,
+          "maximum": 9007199254740991,
+          "type": "integer"
+        },
+        "respondentCount": {
+          "description": "Total respondent rows in the reviewed sheet, copied from the preview.",
+          "maximum": 9007199254740991,
+          "minimum": 0,
+          "type": "integer"
+        },
+        "retainedCombinationIds": {
+          "items": {
+            "type": "string"
+          },
+          "maxItems": 5000,
+          "type": "array"
+        },
+        "segmentationColumns": {
+          "description": "Keys of the variables the user kept after review. These are the only ones that shape the cohort.",
+          "items": {
+            "type": "string"
+          },
+          "maxItems": 200,
+          "minItems": 1,
+          "type": "array"
+        },
+        "sheetName": {
+          "type": "string"
+        },
+        "variables": {
+          "description": "Variable-level distributions from the preview (preferred over `columns` when present).",
+          "items": {
+            "properties": {
+              "distribution": {
+                "items": {
+                  "properties": {
+                    "respondentCount": {
+                      "maximum": 9007199254740991,
+                      "minimum": 0,
+                      "type": "integer"
+                    },
+                    "sharePct": {
+                      "minimum": 0,
+                      "type": "number"
+                    },
+                    "value": {
+                      "default": "",
+                      "type": "string"
+                    }
+                  },
+                  "required": [
+                    "value",
+                    "respondentCount",
+                    "sharePct"
+                  ],
+                  "type": "object"
+                },
+                "maxItems": 30,
+                "type": "array"
+              },
+              "key": {
+                "type": "string"
+              },
+              "label": {
+                "type": "string"
+              },
+              "otherRespondentCount": {
+                "maximum": 9007199254740991,
+                "minimum": 0,
+                "type": "integer"
+              },
+              "validRespondentCount": {
+                "maximum": 9007199254740991,
+                "minimum": 0,
+                "type": "integer"
+              }
+            },
+            "required": [
+              "key",
+              "label",
+              "distribution"
+            ],
+            "type": "object"
+          },
+          "maxItems": 200,
+          "type": "array"
+        }
+      },
+      "required": [
+        "fileName",
+        "segmentationColumns",
+        "combinations"
+      ],
+      "type": "object"
+    },
+    "excludeDistributions": {
+      "description": "Exact axis names whose composition quotas should be removed. Age remains as one eligibility range without within-range shares; explicit gender membership remains enforced. Include required eligibility axes in cohortAllocation.distributionNames. Original evidence stays in reviewDistributions.",
+      "items": {
+        "maxLength": 200,
+        "minLength": 1,
+        "type": "string"
+      },
+      "maxItems": 200,
+      "type": "array"
+    },
+    "memberCount": {
+      "description": "Exact number of Minds to create. Pass this whenever the user states a size (\"exactly 50 per region\", \"genau 50 Minds je Zelle\") instead of relying on the server to parse the number out of the brief. When omitted, the size is inferred from the brief and, failing that, from the creation mode's evidence-based automatic sizing. Two separate refusals apply: 403 MODE_CAP above 20 in \"balanced\" mode (switch to segment_coverage or benchmark_depth), and 403 PLAN_LIMIT above your plan's per-Audience cap. The 6000 accepted here is the absolute system ceiling, not your allowance — call get_audience_limits for the real one. The Audience is never created at a partial size.",
+      "exclusiveMinimum": 0,
+      "maximum": 6000,
+      "type": "integer"
+    }
+  },
+  "type": "object"
+}

Removed value: -{
-  "description": "Reviewed variable roles, distributions, and privacy-safe pairwise relationships returned by preview_audience_dataset_segmentation. Requires audienceCreationMode=\"benchmark_depth\". Structural variables shape one generalizable representative cohort; outcomes remain held out and joint combinations remain audit evidence only. Copy `respondentCount` and `recommendedMindCount` through from the preview — they cap the cohort size.",
-  "properties": {
-    "columns": {
-      "description": "Column-level distributions from the preview. Used when `variables` is absent.",
-      "items": {
-        "properties": {
-          "distribution": {
-            "items": {
-              "properties": {
-                "respondentCount": {
-                  "maximum": 9007199254740991,
-                  "minimum": 0,
-                  "type": "integer"
-                },
-                "sharePct": {
-                  "minimum": 0,
-                  "type": "number"
-                },
-                "value": {
-                  "default": "",
-                  "type": "string"
-                }
-              },
-              "required": [
-                "value",
-                "respondentCount",
-                "sharePct"
-              ],
-              "type": "object"
-            },
-            "maxItems": 30,
-            "type": "array"
-          },
-          "key": {
-            "type": "string"
-          },
-          "label": {
-            "type": "string"
-          },
-          "otherRespondentCount": {
-            "maximum": 9007199254740991,
-            "minimum": 0,
-            "type": "integer"
-          },
-          "validRespondentCount": {
-            "maximum": 9007199254740991,
-            "minimum": 0,
-            "type": "integer"
-          }
-        },
-        "required": [
-          "key",
-          "label",
-          "distribution"
-        ],
-        "type": "object"
-      },
-      "maxItems": 200,
-      "type": "array"
-    },
-    "combinations": {
-      "description": "Observed joint profiles as audit counts. Copy the ids and counts from the preview; per-cell values are not needed.",
-      "items": {
-        "properties": {
-          "id": {
-            "type": "string"
-          },
-          "respondentCount": {
-            "exclusiveMinimum": 0,
-            "maximum": 9007199254740991,
-            "type": "integer"
-          },
-          "sharePct": {
-            "minimum": 0,
-            "type": "number"
-          }
-        },
-        "required": [
-          "id",
-          "respondentCount",
-          "sharePct"
-        ],
-        "type": "object"
-      },
-      "maxItems": 5000,
-      "type": "array"
-    },
-    "fileName": {
-      "type": "string"
-    },
-    "recommendedMindCount": {
-      "description": "Preview-recommended cohort size. Caps how many Minds the benchmark run creates.",
-      "exclusiveMinimum": 0,
-      "maximum": 9007199254740991,
-      "type": "integer"
-    },
-    "respondentCount": {
-      "description": "Total respondent rows in the reviewed sheet, copied from the preview.",
-      "maximum": 9007199254740991,
-      "minimum": 0,
-      "type": "integer"
-    },
-    "retainedCombinationIds": {
-      "items": {
-        "type": "string"
-      },
-      "maxItems": 5000,
-      "type": "array"
-    },
-    "segmentationColumns": {
-      "description": "Keys of the variables the user kept after review. These are the only ones that shape the cohort.",
-      "items": {
-        "type": "string"
-      },
-      "maxItems": 200,
-      "minItems": 1,
-      "type": "array"
-    },
-    "sheetName": {
-      "type": "string"
-    },
-    "variables": {
-      "description": "Variable-level distributions from the preview (preferred over `columns` when present).",
-      "items": {
-        "properties": {
-          "distribution": {
-            "items": {
-              "properties": {
-                "respondentCount": {
-                  "maximum": 9007199254740991,
-                  "minimum": 0,
-                  "type": "integer"
-                },
-                "sharePct": {
-                  "minimum": 0,
-                  "type": "number"
-                },
-                "value": {
-                  "default": "",
-                  "type": "string"
-                }
-              },
-              "required": [
-                "value",
-                "respondentCount",
-                "sharePct"
-              ],
-              "type": "object"
-            },
-            "maxItems": 30,
-            "type": "array"
-          },
-          "key": {
-            "type": "string"
-          },
-          "label": {
-            "type": "string"
-          },
-          "otherRespondentCount": {
-            "maximum": 9007199254740991,
-            "minimum": 0,
-            "type": "integer"
-          },
-          "validRespondentCount": {
-            "maximum": 9007199254740991,
-            "minimum": 0,
-            "type": "integer"
-          }
-        },
-        "required": [
-          "key",
-          "label",
-          "distribution"
-        ],
-        "type": "object"
-      },
-      "maxItems": 200,
-      "type": "array"
-    }
-  },
-  "required": [
-    "fileName",
-    "segmentationColumns",
-    "combinations"
-  ],
-  "type": "object"
-}

Removed value: -{
-  "description": "Exact axis names whose composition quotas should be removed. Age remains as one eligibility range without within-range shares; explicit gender membership remains enforced. Include required eligibility axes in cohortAllocation.distributionNames. Original evidence stays in reviewDistributions.",
-  "items": {
-    "maxLength": 200,
-    "minLength": 1,
-    "type": "string"
-  },
-  "maxItems": 200,
-  "type": "array"
-}

Removed value: -{
-  "description": "Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance.",
-  "items": {
-    "properties": {
-      "name": {
-        "description": "File name",
-        "type": "string"
-      },
-      "url": {
-        "description": "URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset.",
-        "minLength": 1,
-        "type": "string"
-      }
-    },
-    "required": [
-      "name",
-      "url"
-    ],
-    "type": "object"
-  },
-  "maxItems": 20,
-  "type": "array"
-}

Added value: +{
+  "description": "Review the quota axes before any Mind exists, and pass a reviewed snapshot back unchanged to create from it.",
+  "properties": {
+    "preview": {
+      "description": "Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool polls it to completion. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options.",
+      "type": "boolean"
+    },
+    "reviewedJson": {
+      "description": "Exact reviewedGroundingJson returned by groundingPreview or import_audience_sources. Pass it unchanged together with reviewedGroundingSha256 on a re-preview or creation call; the server reuses this reviewed snapshot instead of rerunning grounding research.",
+      "maxLength": 2000000,
+      "minLength": 1,
+      "type": "string"
+    },
+    "reviewedSha256": {
+      "description": "SHA-256 returned by the same groundingPreview. Required together with reviewedGroundingJson; any mismatch is rejected before Audience or Mind materialisation.",
+      "pattern": "^[0-9a-f]{64}$",
+      "type": "string"
+    }
+  },
+  "type": "object"
+}

Removed value: -{
-  "description": "Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool polls it to completion. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options.",
-  "type": "boolean"
-}

Removed value: -{
-  "description": "Legacy alias for audienceCreationMode — same values, same behaviour. How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account.",
-  "enum": [
-    "balanced",
-    "segment_coverage",
-    "benchmark_depth"
-  ],
-  "type": "string"
-}

Removed value: -{
-  "default": true,
-  "description": "Set false to skip Exa web search and extraction completely. The Audience is then grounded only on the brief and supplied files, links, keywords, and structurally parsed respondent data.",
-  "type": "boolean"
-}

Removed value: -{
-  "description": "Optional Exa web-search seeds added alongside the brief.",
-  "items": {
-    "type": "string"
-  },
-  "maxItems": 10,
-  "type": "array"
-}

Removed value: -{
-  "description": "Optional URLs scraped server-side for additional context (e.g. an article describing the population).",
-  "items": {
-    "type": "string"
-  },
-  "maxItems": 10,
-  "type": "array"
-}

Removed value: -{
-  "description": "Exact number of Minds to create. Pass this whenever the user states a size (\"exactly 50 per region\", \"genau 50 Minds je Zelle\") instead of relying on the server to parse the number out of the brief. When omitted, the size is inferred from the brief and, failing that, from the creation mode's evidence-based automatic sizing. Two separate refusals apply: 403 MODE_CAP above 20 in \"balanced\" mode (switch to segment_coverage or benchmark_depth), and 403 PLAN_LIMIT above your plan's per-Audience cap. The 6000 accepted here is the absolute system ceiling, not your allowance — call get_audience_limits for the real one. The Audience is never created at a partial size.",
-  "exclusiveMinimum": 0,
-  "maximum": 6000,
-  "type": "integer"
-}

Added value: +{
+  "description": "Extra material the grounding research reads alongside the brief, and whether web search runs at all.",
+  "properties": {
+    "files": {
+      "description": "Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance.",
+      "items": {
+        "properties": {
+          "name": {
+            "description": "File name",
+            "type": "string"
+          },
+          "url": {
+            "description": "URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset.",
+            "minLength": 1,
+            "type": "string"
+          }
+        },
+        "required": [
+          "name",
+          "url"
+        ],
+        "type": "object"
+      },
+      "maxItems": 20,
+      "type": "array"
+    },
+    "includeWebSearch": {
+      "default": true,
+      "description": "Set false to skip Exa web search and extraction completely. The Audience is then grounded only on the brief and supplied files, links, keywords, and structurally parsed respondent data.",
+      "type": "boolean"
+    },
+    "keywords": {
+      "description": "Optional Exa web-search seeds added alongside the brief.",
+      "items": {
+        "type": "string"
+      },
+      "maxItems": 10,
+      "type": "array"
+    },
+    "links": {
+      "description": "Optional URLs scraped server-side for additional context (e.g. an article describing the population).",
+      "items": {
+        "type": "string"
+      },
+      "maxItems": 10,
+      "type": "array"
+    }
+  },
+  "type": "object"
+}

Removed value: -{
-  "description": "Exact reviewedGroundingJson returned by groundingPreview or import_audience_sources. Pass it unchanged together with reviewedGroundingSha256 on a re-preview or creation call; the server reuses this reviewed snapshot instead of rerunning grounding research.",
-  "maxLength": 2000000,
-  "minLength": 1,
-  "type": "string"
-}

Removed value: -{
-  "description": "SHA-256 returned by the same groundingPreview. Required together with reviewedGroundingJson; any mismatch is rejected before Audience or Mind materialisation.",
-  "pattern": "^[0-9a-f]{64}$",
-  "type": "string"
-}

Removed value: -{
-  "description": "Legacy alias for `brief`. Accepted for back-compat.",
-  "maxLength": 200000,
-  "minLength": 1,
-  "type": "string"
-}

Removed value: -{
-  "description": "Deprecated; omit it. Every member of every Audience is trained individually in the background, including reviewed-dataset cohorts of more than ~40 Minds (each keeps its deterministic cohort profile and the request locale). true is accepted as a no-op; false is rejected, because it used to create untrained Minds with an empty system prompt that reported ready. Cost and time: one Mind training per member; a 100-Mind cohort takes tens of minutes to a few hours. The tool returns as soon as the Audience exists (structuredContent.audience.memberTraining reports what was queued); wait until every member reports readyToChat (get_mind_status or GET /api/v1/minds/{id}/training) before running a Study, which refuses Audiences with untrained members.",
-  "type": "boolean"
-}

Previous value: -nullNew value: +{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "additionalProperties": {},
+  "properties": {
+    "audience": {
+      "description": "The created or previewed Audience with its Minds and grounding."
+    },
+    "operation": {
+      "description": "Durable creation operation, while creation is still running."
+    },
+    "status": {
+      "description": "Creation state of the Audience."
+    },
+    "workspaceUrl": {
+      "anyOf": [
+        {
+          "type": "string"
+        },
+        {
+          "type": "null"
+        }
+      ],
+      "description": "Minds workspace link for the Audience."
+    }
+  },
+  "type": "object"
+}

Previous value: -"Opt-in per-Mind training for large reviewed-dataset cohorts (default false). By default a reviewed-dataset Audience of more than ~40 Minds is bulk-created immediately ready: exact demographics, but no per-Mind research training (empty system prompt, no knowledge items). Set true to create those Minds untrained instead and enqueue each one through the same per-member training pipeline the in-app draft flow uses, preserving its deterministic cohort profile and the request locale. The tool still returns as soon as the Audience exists — training continues in the background (structuredContent.audience.memberTraining reports what was queued); poll each member with GET /api/v1/minds/{id}/training. Has no effect on Audiences whose members are already trained individually."New value: +"Deprecated; omit it. Every member of every Audience is trained individually in the background, including reviewed-dataset cohorts of more than ~40 Minds (each keeps its deterministic cohort profile and the request locale). true is accepted as a no-op; false is rejected, because it used to create untrained Minds with an empty system prompt that reported ready. Cost and time: one Mind training per member; a 100-Mind cohort takes tens of minutes to a few hours. The tool returns as soon as the Audience exists (structuredContent.audience.memberTraining reports what was queued); wait until every member reports readyToChat (get_mind_status or GET /api/v1/minds/{id}/training) before running a Study, which refuses Audiences with untrained members."

Previous value: -4000New value: +200000

Previous value: -3New value: +1

Previous value: -"Exact axis names (copied from a preview) that must not be used as cohort quotas. Applied deterministically after grounding; the evidence stays visible in reviewDistributions."New value: +"Exact axis names whose composition quotas should be removed. Age remains as one eligibility range without within-range shares; explicit gender membership remains enforced. Include required eligibility axes in cohortAllocation.distributionNames. Original evidence stays in reviewDistributions."

Previous value: -"Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."New value: +"Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."

Previous value: -"URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. The Study refuses to start if Minds cannot read the asset."New value: +"URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. A workspace upload reused as a Study asset must sit under chat/<userId>/: request the signed upload with folder \"chat\" and pass that storage path or its /api/uploads/chat/ URL. A temp/, portfolio/, or context/ path — and any /api/uploads/file-access/ URL — is refused as not belonging to the workspace owner and is never re-fetched; pass a plain external URL instead. The Study refuses to start if Minds cannot read the asset."

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. Fresh research uses the same durable operation as the public REST API; this tool polls it to completion. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."

Previous value: -4000New value: +200000

Previous value: -3New value: +1

Previous value: -200New value: +500

Previous value: -200New value: +500

Previous value: -50New value: +200

Previous value: -200New value: +500

Previous value: -10New value: +20

Previous value: -"Exact reviewedGroundingJson returned by a prior groundingPreview. Pass it unchanged together with reviewedGroundingSha256 on a re-preview or creation call; the server reuses this reviewed snapshot instead of rerunning grounding research."New value: +"Exact reviewedGroundingJson returned by groundingPreview or import_audience_sources. Pass it unchanged together with reviewedGroundingSha256 on a re-preview or creation call; the server reuses this reviewed snapshot instead of rerunning grounding research."

Previous value: -250000New value: +2000000

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. To revise a saved preview without repeating research, pass its unchanged reviewedGroundingJson/SHA pair together with groundingPreview and the revised exclusion/allocation options. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."

Previous value: -"Exact reviewedGroundingJson returned by a prior groundingPreview. Pass it unchanged together with reviewedGroundingSha256 on the creation call; the server reuses this reviewed snapshot instead of rerunning grounding research."New value: +"Exact reviewedGroundingJson returned by a prior groundingPreview. Pass it unchanged together with reviewedGroundingSha256 on a re-preview or creation call; the server reuses this reviewed snapshot instead of rerunning grounding research."

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and returns the axes it would allocate on — with Study-outcome axes demoted and listed in flaggedDistributions — without creating an Audience or Minds and without consuming a generation allowance. Present the axes to the user for approval, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged so creation uses the reviewed snapshot instead of rerunning research."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and deterministic cohort allocation without creating an Audience or Minds or consuming a generation allowance. The response includes effective target percentages and integer counts; explicit allocation targets that do not match grounded axes fail closed. Present the plan to the user, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged with the same allocation options."

Added value: +{
+  "description": "Resume a queued preview or creation by its returned jobId. Reads its v1 operation status without creating or charging anything. Supply this alone; do not repeat the creation brief to check progress.",
+  "format": "uuid",
+  "pattern": "^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$",
+  "type": "string"
+}

Previous value: -"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and returns the axes it would allocate on — with Study-outcome axes demoted and listed in flaggedDistributions — without creating an Audience or Minds and without consuming a generation allowance. Present the axes to the user for approval, then call again without groundingPreview (same brief and options) to create, passing any axis names to drop as excludeDistributions."New value: +"Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and returns the axes it would allocate on — with Study-outcome axes demoted and listed in flaggedDistributions — without creating an Audience or Minds and without consuming a generation allowance. Present the axes to the user for approval, then call again without groundingPreview and pass reviewedGroundingJson plus reviewedGroundingSha256 unchanged so creation uses the reviewed snapshot instead of rerunning research."

Added value: +{
+  "description": "Exact reviewedGroundingJson returned by a prior groundingPreview. Pass it unchanged together with reviewedGroundingSha256 on the creation call; the server reuses this reviewed snapshot instead of rerunning grounding research.",
+  "maxLength": 250000,
+  "minLength": 1,
+  "type": "string"
+}

Added value: +{
+  "description": "SHA-256 returned by the same groundingPreview. Required together with reviewedGroundingJson; any mismatch is rejected before Audience or Mind materialisation.",
+  "pattern": "^[0-9a-f]{64}$",
+  "type": "string"
+}

Added value: +{
+  "description": "Exact axis names (copied from a preview) that must not be used as cohort quotas. Applied deterministically after grounding; the evidence stays visible in reviewDistributions.",
+  "items": {
+    "maxLength": 200,
+    "minLength": 1,
+    "type": "string"
+  },
+  "maxItems": 200,
+  "type": "array"
+}

Added value: +{
+  "description": "Review the quota axes BEFORE any Mind exists. When true, the server runs research grounding and returns the axes it would allocate on — with Study-outcome axes demoted and listed in flaggedDistributions — without creating an Audience or Minds and without consuming a generation allowance. Present the axes to the user for approval, then call again without groundingPreview (same brief and options) to create, passing any axis names to drop as excludeDistributions.",
+  "type": "boolean"
+}

Previous value: -"Audience creation mode. Preferred. Use balanced for everyday research, segment_coverage for coverage of limiting grounded cells, or benchmark_depth for validation-ready repeated representation."New value: +"How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account."

Previous value: -"Audience creation mode. \"balanced\" (default): compact, representative Audience for everyday research. \"segment_coverage\": two representatives per limiting grounded audience cell, with a 10-Mind evidence floor and 50-Mind mode ceiling. \"benchmark_depth\": repeated representation of limiting grounded audience cells for validation-ready benchmark / survey-style segmentation, with the paid allowance as its ceiling. The deeper modes require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in the response (structuredContent.audience.audienceCreationMode). This field is the legacy alias for audienceCreationMode."New value: +"Legacy alias for audienceCreationMode — same values, same behaviour. How the Audience is sized. \"balanced\" (default): compact and representative for everyday research. It is the only mode with a hard wall on a size you state — a memberCount above 20 is refused with 403 MODE_CAP. \"segment_coverage\": two representatives per limiting grounded cell, from a 10-Mind floor; automatic sizing stops at 50, but a memberCount you state is bounded only by your plan. \"benchmark_depth\": repeated representation of limiting grounded cells for validation-ready segmentation; bounded by your plan allowance. The deeper two require a Team (enterprise) plan — on other plans the server SILENTLY downgrades to \"balanced\" and echoes the effective mode in structuredContent.audience.audienceCreationMode. Call get_audience_limits to see the ceilings that apply to your account."

Previous value: -"Exact number of Minds to create in the Audience. Pass this whenever the user states a size (\"exactly 50 per region\", \"genau 50 Minds je Zelle\") instead of relying on the server to parse the number out of the brief prose. When omitted, the size is inferred from the brief and, failing that, from the creation mode's evidence-based automatic sizing. Rejected with 403 PLAN_LIMIT when it exceeds the plan's per-Audience cap; the Audience is never created at a partial size."New value: +"Exact number of Minds to create. Pass this whenever the user states a size (\"exactly 50 per region\", \"genau 50 Minds je Zelle\") instead of relying on the server to parse the number out of the brief. When omitted, the size is inferred from the brief and, failing that, from the creation mode's evidence-based automatic sizing. Two separate refusals apply: 403 MODE_CAP above 20 in \"balanced\" mode (switch to segment_coverage or benchmark_depth), and 403 PLAN_LIMIT above your plan's per-Audience cap. The 6000 accepted here is the absolute system ceiling, not your allowance — call get_audience_limits for the real one. The Audience is never created at a partial size."

Previous value: -9007199254740991New value: +6000

Previous value: -"Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."New value: +"Optional already-uploaded research files. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. The Study refuses to start if Minds cannot read the asset. The server analyzes these through the same extended screener/questionnaire path as the in-app New Audience uploader, including study roles, screening and quota rules, review distributions, and grounding provenance."

Previous value: -"URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. The Study refuses to start if Minds cannot read the asset."New value: +"URL of an already-uploaded file. MCP cannot read or upload a local file:// path. Use a fetchable HTTP(S) URL, a signed URL supplied by the client for the attached file, or an existing Minds workspace upload URL/path. Study tools import external file URLs into durable Minds storage before saving or running. The Study refuses to start if Minds cannot read the asset."

Added value: +{
+  "description": "Opt-in per-Mind training for large reviewed-dataset cohorts (default false). By default a reviewed-dataset Audience of more than ~40 Minds is bulk-created immediately ready: exact demographics, but no per-Mind research training (empty system prompt, no knowledge items). Set true to create those Minds untrained instead and enqueue each one through the same per-member training pipeline the in-app draft flow uses, preserving its deterministic cohort profile and the request locale. The tool still returns as soon as the Audience exists — training continues in the background (structuredContent.audience.memberTraining reports what was queued); poll each member with GET /api/v1/minds/{id}/training. Has no effect on Audiences whose members are already trained individually.",
+  "type": "boolean"
+}