Ballet
Server Details
Ballet is AI workflow automation for RevOps, CS Ops and Marketing Ops (MarkOps). Build and run ops workflows from Claude or ChatGPT: CRM hygiene, lead routing and enrichment, renewals, churn alerts, handoffs and approvals, with humans in the loop.
- Status
- Healthy
- Uptime
- 25.0% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- Ballet
TDQS
Scored across 86 tools
The tools cover many resources and lifecycle steps, but each action generally has a distinct target and descriptions include explicit 'use when' guidance. Some near-neighbors around report panels, record update vs upsert, and trigger lifecycle could still be confused without reading the detailed descriptions.
All tool names use snake_case with predictable verb_noun phrasing such as list_*, get_*, create_*, update_*, delete_*, publish_*, and run_*. Minor verb variation like add, remove, move, upsert, and try is appropriate and readable.
With 86 tools, the surface is far beyond a well-scoped MCP server and crosses the rubric's 50+ extreme-mismatch threshold. The platform domain is broad, but the set should be split into smaller focused servers or substantially trimmed.
Coverage is unusually broad: CRUD for agents, playbooks, reports and panels, data tables and records, triggers, MCP servers, secrets, skills, and custom tools, plus runs, analytics, guides, models, and integrations. Some gaps remain, such as no delete_trigger, no collection update/delete, and no run cancellation, but core workflows are covered.
Available Tools
86 toolsadd_panelAInspect
Append one panel to a report without resending the others. The panel needs a unique id, layout, presentation fields, and a query. Saves one report version (max 24 panels). Use when extending an existing dashboard.
Inputs: panel (required), reportId (required), expectedVersion.
connect_mcp_serverAIdempotentInspect
Connect or reconnect to an existing MCP server by ID. Use when a connector is configured but not currently connected.
Inputs: id (required).
connect_playbook_to_triggerADestructiveIdempotentInspect
Make a playbook use a trigger: it gets the items the trigger finds that its filter says yes to. Saves on the playbook's working copy; it only takes effect when the playbook is published, and that needs the trigger published first. rule says in plain words which items it runs for; filter is a module with export function filter(record) (it must say export, or it fails on every item). try_playbook_filter can validate the filter first. Connecting again to the same trigger replaces the filter.
Inputs: rule (required), filter (required), triggerId (required), playbookId (required).
create_agentAInspect
Create a new AI agent in the workspace. Use when a playbook step needs LLM reasoning (for example ticket classification or lead scoring). Model should use provider "gateway" and a model id from list_models. Optional skills[] and tools[] attach tenant skills and HTTP tools. Authoring details are in get_builder_guide topic "agent".
Inputs: name (required), model (required), instructions (required), tools, +2 more.
create_collectionAInspect
Create a collection (a group of playbooks). Use when none of the existing collections from list_collections fit a new automation. Returns the collection including its id for create_playbook. name is required; description and color are optional. Slug is derived server-side. The new collection is shared with All Members.
Inputs: name (required), color, description.
create_data_tableAInspect
Create a Persistent Data Table with a typed schema (string | number | boolean | datetime). Optional queueConfig seeds reserved queue fields. Returns the table including its id for record operations. Use when a playbook needs structured storage, for example lead lists, invoice queues, or triage inboxes.
Inputs: name (required), fields (required), queueConfig.
create_data_table_recordAInspect
Create a record. Fails if recordKey already exists (use upsert_data_table_record to overwrite). data keys must match declared table fields. Starts at version 1. Use when inserting a new row such as a lead or ticket item.
Inputs: data (required), recordKey (required), dataTableId (required).
create_mcp_serverAInspect
Add a new outbound MCP server configuration and attempt to connect. Use when the user wants Ballet playbooks to call tools from another MCP server.
Inputs: url (required), name (required), transport (required), headers.
create_playbookAInspect
Create a new multi-step playbook in the user's Ballet workspace. Use when the user wants to build a reusable automation, for example a recurring report, lead enrichment flow, invoice approval, or support triage workflow. Requires collectionId from list_collections (or create_collection). Each code step needs id, language, code, input mappings, output.variable (assign that variable, never a top-level return), and sandbox (allowNetwork/secrets/packages). Authoring details are in get_builder_guide topics "playbook" and "sdk".
Inputs: name (required), steps (required), collectionId (required), description, +1 more.
create_reportAInspect
Create a workspace report from a title, defaultRangeDays and panels. A report may start with no panels; viewers don't see it until it has one. Each panel needs a unique id, layout, presentation fields, and a query ES module. Panel type schemas come from list_panel_types. Use when building an ops dashboard such as a weekly report.
Inputs: title (required), panels (required), defaultRangeDays (required).
create_secretAInspect
Store a new workspace secret (UPPER_SNAKE_CASE name). Fails if the name already exists; use update_secret to rotate. The value is never returned. Use when a playbook or tool needs a new API key or credential the vault does not already hold.
Inputs: name (required), value (required), description.
create_skillAInspect
Create a tenant custom skill (runtime instructions agents can pin via skills[]). Use when encoding reusable domain guidance for agents. This is not an authoring guide; those are list_builder_guides / get_builder_guide. Authoring details are in get_builder_guide topic "skill".
Inputs: id (required), body (required), name (required), description (required), +1 more.
create_toolAInspect
Create a custom HTTP tool for APIs that are not registered native integrations. Use when a playbook needs a reusable HTTP call. Auth header values should use "{{SECRET_NAME}}" placeholders from create_secret. Registered vendors must use metaphor.integration in step code instead (see list_integrations). Authoring details are in get_builder_guide topic "tool".
Inputs: id (required), name (required), execute (required), description (required), +2 more.
create_triggerAInspect
Create a trigger: it starts unpublished (nothing runs) and will check every 5 minutes once published. Reuse an existing one when it fits (see list_triggers). scope is what every playbook on it shares, in plain words; fetch is the code that reads the app (contract in get_builder_guide topic "triggers"). After creating, test_trigger then publish_trigger.
Inputs: name (required), event (required), fetch (required), scope (required), +1 more.
delete_agentADestructiveInspect
Permanently delete an agent by ID. Refused while any playbook still references the agent. Use when removing an unused agent.
Inputs: id (required).
delete_data_tableADestructiveInspect
Permanently delete a Persistent Data Table and all of its records. Use when the user no longer needs the table.
Inputs: id (required).
delete_data_table_fieldADestructiveInspect
Delete a field from a Persistent Data Table. Strips that key from every record and increments each record version. Refuses to delete the last field or reserved queue fields. Use when removing a column permanently.
Inputs: id (required), fieldId (required).
delete_data_table_recordADestructiveInspect
Permanently delete one record by recordKey. Use when removing a row the user no longer needs.
Inputs: recordKey (required), dataTableId (required).
delete_mcp_serverADestructiveInspect
Remove an MCP server configuration and disconnect. Use when the user no longer wants that outbound connector.
Inputs: id (required).
delete_playbookADestructiveInspect
Permanently delete a playbook by ID, including its schedules and runs. Use when the user wants to remove an automation they no longer need.
Inputs: id (required).
delete_reportADestructiveInspect
Permanently delete a report. It disappears from list_reports and related tools; there is no undo from MCP. Use when the user no longer needs the dashboard.
Inputs: id (required).
delete_skillADestructiveInspect
Permanently delete a custom skill by ID. Use when removing unused tenant skill content.
Inputs: id (required).
delete_toolADestructiveInspect
Permanently delete a custom HTTP tool by ID. Use when removing an unused tool definition.
Inputs: id (required).
disconnect_playbook_from_triggerADestructiveIdempotentInspect
Stop a playbook using a trigger. Saves on the playbook's working copy; it takes effect when the playbook is published, and until then the trigger keeps sending it items. When it takes effect, items still waiting for that playbook are closed as "never ran". A trigger can only be deleted once no published playbook uses it.
Inputs: triggerId (required), playbookId (required).
get_agentARead-onlyIdempotentInspect
Get an agent by ID, including instructions, model, tools, and skills. Use when inspecting or editing an existing agent.
Inputs: id (required).
get_builder_guideARead-onlyIdempotentInspect
Return full markdown for Ballet authoring guides (topics such as overview, playbook, sdk, agent, tool, skill, connector, report). Use when building or editing the matching artifact so step shapes and SDK contracts match Ballet. Guide ids come from list_builder_guides.
Inputs: topics (required).
get_data_tableARead-onlyIdempotentInspect
Get one Persistent Data Table by id, including its typed field schema. Use when inspecting or editing a table before working with records.
Inputs: id (required).
get_data_table_recordARead-onlyIdempotentInspect
Get one record by recordKey. Response includes version for optimistic locking on update_data_table_record. Use when reading a single row before updating it.
Inputs: recordKey (required), dataTableId (required).
get_mcp_serverARead-onlyIdempotentInspect
Get an MCP server by ID with live connection status. Use when inspecting or troubleshooting a connector.
Inputs: id (required).
get_panel_dataARead-onlyIdempotentInspect
Get one saved panel's data for a date range (the same numbers the report page shows). from/to are inclusive YYYY-MM-DD; omit both for the report default range. Returns { status: "ok", data, range } or { status: "error" }. Use when reading metrics from an existing weekly or ops report.
Inputs: panelId (required), reportId (required), to, from.
get_playbookARead-onlyIdempotentInspect
Get a playbook by ID, including its steps and configuration. Use when inspecting how an automation works before editing or running it. The result may include studioUrl for opening the playbook in Studio.
Inputs: id (required).
get_reportARead-onlyIdempotentInspect
Get one workspace report by id. Includes panel queries for Staff MCP and editors with report authoring enabled; otherwise queries are omitted. Use when inspecting a dashboard definition.
Inputs: id (required).
get_runARead-onlyIdempotentInspect
Get the status and result of a playbook run. Use after run_playbook, or when the user asks whether an automation finished and what it produced.
Inputs: id (required).
get_run_stepsARead-onlyIdempotentInspect
Get per-step results for a playbook run. Use when diagnosing a failed or unexpected run.
Inputs: id (required).
get_skillARead-onlyIdempotentInspect
Get a skill by ID, including its full instruction body. Use when reviewing or editing a skill before attaching it to an agent.
Inputs: id (required).
get_toolARead-onlyIdempotentInspect
Get a custom HTTP tool by ID (secrets masked). Use when inspecting or editing an existing tool definition.
Inputs: id (required).
get_triggerARead-onlyIdempotentInspect
Get one trigger in full: its working copy (app, event, scope, fetch code), the playbooks that use it with their rules and filter code, its schedule and last check, and the result of its last test. It also gives definitionHash and lookBackTarget, which publish_trigger needs back. Use when inspecting a trigger or confirming what would go live before publish_trigger.
Inputs: triggerId (required), timeZone.
get_workspace_daily_trendsARead-onlyIdempotentInspect
Daily activity for this workspace over a date range (default last 30 days, max 366): run volume and outcomes, active members, console sessions, MCP calls, playbook and skill executions, and LLM tokens. Only days with activity are returned; treat missing days as zero. Use for usage trends; use get_workspace_roi for time/value saved.
Inputs: to, from.
get_workspace_healthARead-onlyIdempotentInspect
Current workspace adoption health: score (0-100), band (healthy | watch | at_risk), and trailing 30-day signals. health is null until the daily rollup has run. Use when checking whether the workspace is actively using automations.
Inputs: no named parameters.
get_workspace_roiARead-onlyIdempotentInspect
Estimated time and labor value saved by playbook runs in this workspace over a date range (default last 30 days, max 366). Returns totals and, unless includePlaybooks is false, a per-playbook breakdown sorted by minutes saved. Only successful runs count. Money is in USD cents. Use when the user asks how much automation is saving.
Inputs: to, from, includePlaybooks.
list_agentsARead-onlyIdempotentInspect
List AI agents in the workspace. Use when finding an agent to reuse in a playbook step or before create_agent.
Inputs: no named parameters.
list_builder_guidesARead-onlyIdempotentInspect
List Ballet authoring guides available to this workspace (id, name, description, keywords). Use when choosing which get_builder_guide topics to load before creating a playbook, agent, tool, or skill.
Inputs: query.
list_collectionsARead-onlyIdempotentInspect
List collections in the workspace (groups of playbooks). Use when choosing a collectionId for create_playbook or organizing automations.
Inputs: no named parameters.
list_data_table_recordsARead-onlyIdempotentInspect
List or query records in a Persistent Data Table. Supports filter, sort, limit/take (max 100), skip, cursor, and search. Use when reading rows for reporting, enrichment, or triage workflows. skip/take cannot combine with cursor; cursor pagination needs the default createdAt sort.
Inputs: dataTableId (required), skip, sort, take, +4 more.
list_data_tablesARead-onlyIdempotentInspect
List Persistent Data Tables in the workspace (id, name, fields). Use when finding a table id for record tools; never invent a display name as the table id.
Inputs: no named parameters.
list_integrationsARead-onlyIdempotentInspect
List registered native integrations for this workspace. Unscoped calls return id, name, and operationIds; pass ids or query (for example { ids: ["zendesk"] }) for operation descriptions plus requestSchema/responseSchema. Use when wiring a playbook to a vendor (CRM, helpdesk, email) so step code can call metaphor.integration instead of inventing a custom HTTP tool.
Inputs: ids, query, includeOperations.
list_mcp_serversARead-onlyIdempotentInspect
List MCP servers connected to this Ballet workspace and their connection status. Use when discovering outbound connectors the playbooks can call.
Inputs: no named parameters.
list_mcp_toolsARead-onlyIdempotentInspect
List tools available from a connected outbound MCP server. Requires serverId from list_mcp_servers. Use when discovering what that connector can do.
Inputs: serverId (required).
list_modelsARead-onlyIdempotentInspect
List LLM and related models available to this workspace (current Fast / Balanced / Capable per provider, plus evaluation models). Omit modality for language and evaluation; pass "evaluation", "image", "video", or "all" to narrow. Use when choosing a model for create_agent or update_agent (provider "gateway", model id from this list).
Inputs: modality.
list_panel_typesARead-onlyIdempotentInspect
List the panel types a workspace report can contain, with presentation and output JSON Schemas. Use when authoring report panels for ops dashboards such as weekly run volume or ticket throughput.
Inputs: no named parameters.
list_playbooksARead-onlyIdempotentInspect
List playbooks in the user's Ballet workspace. Use when the user wants to find, choose, or inspect an existing automation before running or editing it. Optionally filter by collectionId.
Inputs: collectionId.
list_reportsARead-onlyIdempotentInspect
List workspace reports (id, title, updatedAt). Use when finding a report before reading panels or editing it.
Inputs: no named parameters.
list_report_versionsARead-onlyIdempotentInspect
List a report's saved versions, newest first (version, when, who, source tool, summary, restoredFrom). Use when choosing a version to restore with restore_report_version.
Inputs: reportId (required).
list_runsARead-onlyIdempotentInspect
List recent playbook runs in the workspace, optionally filtered by playbookId. Use when reviewing run history for an automation or the workspace.
Inputs: limit, playbookId.
list_secretsARead-onlyIdempotentInspect
List secret names stored for this workspace (name, description, updatedAt). Values are never returned. Use when checking whether credentials for an integration or HTTP tool already exist before create_secret or update_secret.
Inputs: no named parameters.
list_skillsARead-onlyIdempotentInspect
List skills available to agents (built-in and tenant custom). Optionally filter by keyword query. Use when choosing skills to pin on an agent.
Inputs: query.
list_toolsARead-onlyIdempotentInspect
List custom HTTP tools in the workspace (secrets masked). Use when finding reusable HTTP actions for playbooks or agents. Native integrations are listed separately by list_integrations.
Inputs: no named parameters.
list_triggersARead-onlyIdempotentInspect
List the triggers in this workspace: each watches one app for one kind of change and hands what it finds to the playbooks that use it. Use before creating one, to reuse an existing trigger when it already fits. Each says if it is published, whether a publish still needs a test, which playbooks use it, when it runs and what its last check did. Authoring details are in get_builder_guide topic "triggers".
Inputs: timeZone.
move_panelsAInspect
Change the layout of one or more panels in a single report version without touching queries or presentation. Use when rearranging a dashboard.
Inputs: moves (required), reportId (required), expectedVersion.
pause_triggerAIdempotentInspect
Pause a trigger: it stops checking at once and keeps where it got to. The playbooks that use it get nothing new until it is resumed. Use when stopping a misbehaving trigger.
Inputs: triggerId (required).
preview_panelAInspect
Run a draft panel query against this workspace's real data for a date range without saving. Returns { status: "ok", data } or { status: "error", details }. from/to are inclusive YYYY-MM-DD, at most 366 days apart. Use when iterating on a panel before create_report or update_report.
Inputs: to (required), from (required), panel (required).
publish_playbookADestructiveInspect
Publish a playbook: snapshot the working copy and promote it to the live version. Webhooks and run_playbook use the live version. Use after editing when the user wants the new definition to be what runs. Fails if another member holds the edit lock.
Inputs: id (required).
publish_triggerADestructiveInspect
Make the trigger's working copy live: it starts checking the app on its schedule and handing items to the playbooks that use it. Refused unless its last test was of exactly this working copy and passed. A trigger can be shared by several playbooks, so confirm with the user what goes live before calling this. Pass definitionHash from get_trigger as expectedDefinitionHash so publish is refused if the trigger changed while deciding. cursor says where it carries on from: { to: "keep" } (default), { to: "now" }, { to: "hoursBack", hours } (up to 30 days) or { to: "time", time }. Anything but keep looks at older items again: what a playbook already handled won't run twice, but what it hasn't handled yet will, so confirm first and pass lookBackTarget from the same get_trigger as expectedLookBackTarget for hoursBack or time. A playbook can only be published once the triggers it uses are published.
Inputs: triggerId (required), expectedDefinitionHash (required), cursor, expectedLookBackTarget.
remove_panelADestructiveInspect
Remove one panel by id and save a report version. Use when deleting a chart or KPI from a dashboard.
Inputs: panelId (required), reportId (required), expectedVersion.
rename_data_table_fieldAInspect
Rename a field on a Persistent Data Table. Rewrites matching JSON keys on every record and increments each record version. Refuses reserved queue fields on tables with queueConfig. Use when changing a column name without deleting data.
Inputs: id (required), name (required), fieldId (required).
request_playbook_setupAInspect
Create a setup request for secrets and connections a playbook still needs, without collecting secret values in chat. Never ask the user to paste a secret value into chat. Use when verify_playbook or run_scratch_code reports missing secrets or disconnected integrations or MCP connectors. Without recipientEmail, returns setupUrl for the user to complete in the browser. With recipientEmail, emails that person the link instead and does not return the URL. Returns { needed: false } when nothing is missing. Links expire after 48 hours; a new link without recipientEmail invalidates the previous one.
Inputs: playbookId (required), recipientEmail.
restore_report_versionADestructiveInspect
Roll a report back to an earlier version (title, default range, panels including queries). Saved as a new version so history is kept. Use when undoing a bad dashboard edit.
Inputs: version (required), reportId (required), expectedVersion.
resume_triggerADestructiveIdempotentInspect
Resume a paused trigger: it starts checking again on its schedule, and the playbooks that use it start getting items again. Use when re-enabling a paused trigger after confirming with the user.
Inputs: triggerId (required).
run_playbookADestructiveInspect
Start a run of an existing Ballet playbook and return a run ID. Use when the user wants to execute a saved automation such as a weekly report, lead enrichment, invoice approval, or ticket triage. Requires a playbook ID from list_playbooks. Pass wait=true to block up to 50s for the final output; otherwise poll with get_run.
Inputs: id (required), wait, input.
run_scratch_codeAInspect
Run JavaScript or TypeScript in the workspace sandbox without creating a playbook or run record. Use when probing an API, checking credentials, or trying step code before wiring a playbook (for example lead enrichment or invoice lookups). Pass a short plain-English description (3-8 words) shown as the step title. Default and max wait is 50s (timeoutClamped: true if a larger timeout is passed). Default allowNetwork is false; set true for fetch and metaphor.*. Pass secretNames to inject vault values; missingSecret lists names that were absent. Never print secret values. Optional packages install in the sandbox; if install plus execute exceeds 50s the call times out. Existing attachment keys: pass in input and read with metaphor.attachment(key). No upload on this tool.
Inputs: code (required), language (required), input, timeout, +4 more.
set_trigger_scheduleADestructiveIdempotentInspect
Change how often the trigger checks, from the choices the Studio's schedule form has: every N minutes or hours (frequency "custom", with every and unit; N must split the hour or day evenly), "daily", "weekly" (daysOfWeek 0 = Sunday to 6 = Saturday) or "monthly" (day 1 to 28). Give times as the user's own clock (hour 0-23, minute); never convert to UTC, and pass timeZone for a time of day. It takes effect at once, with no publish. It changes how often every playbook that uses the trigger gets items, and how often the app is called: when more than one playbook uses it, confirm with the user first. The answer lists when it will really run next.
Inputs: frequency (required), triggerId (required), hour, unit, +5 more.
test_triggerAInspect
Test the trigger's working copy on recent real records, run every playbook's filter on them, and compare with the live version. Required before publish_trigger, and it counts only for exactly this working copy. Reads only: nothing is queued and nothing runs. The result keeps counts and problems, never the records. madeUpRecords are extra records you write (shaped like the app's) to check that filters survive missing fields; they never count as a pass. Use when validating fetch/filter behaviour before going live.
Inputs: triggerId (required), madeUpRecords.
try_playbook_filterAInspect
Try filter code on a trigger's recent records, before a playbook is connected to it (the trigger's own test only runs the filters of playbooks that are already live). Says how many records each filter said yes to, how many it threw on, and a few matching record keys to look up. Saves nothing and never affects publishing. Uses the trigger's working copy, so it works before the trigger is published.
Inputs: filters (required), triggerId (required).
update_agentBInspect
Update an existing agent by merging the provided fields. Use when changing instructions, model, tools, or skills.
Inputs: id (required), name, model, tools, +3 more.
update_data_tableAInspect
Update a Persistent Data Table: rename, addFields, fieldOrder, updateFields (description only), and/or queueConfig. Type change is not supported. Use rename_data_table_field / delete_data_table_field for column rename and delete. At least one of name, addFields, fieldOrder, updateFields, or queueConfig is required. Setting queueConfig on an empty non-queue table seeds reserved queue fields; refused if the table already has rows.
Inputs: id (required), name, addFields, fieldOrder, +2 more.
update_data_table_recordAInspect
Partial update (merge) of record data. Requires version matching the current record; stale version returns RECORD_CONFLICT with current. Use when safely editing fields under concurrent writers.
Inputs: data (required), version (required), recordKey (required), dataTableId (required).
update_mcp_serverAInspect
Update an MCP server configuration and reconnect. Use when changing URL, transport, headers, or name of an outbound connector.
Inputs: id (required), url, name, headers, +1 more.
update_panelAInspect
Replace one panel, matched by panel.id, without resending the others. Send the whole panel. Saves one report version. Use when editing a single chart or KPI.
Inputs: panel (required), reportId (required), expectedVersion.
update_playbookAInspect
Update an existing playbook by merging the provided fields with the current config. Use when changing steps, name, description, or inputSchema of an automation such as a weekly report or ticket triage flow. The result may include studioUrl.
Inputs: id (required), name, steps, description, +1 more.
update_reportAInspect
Replace fields on an existing report. Pass id plus any of title, defaultRangeDays or panels; omitted fields are kept. panels replaces the full panel list when provided. Use when reshaping a dashboard in one save.
Inputs: id (required), title, panels, expectedVersion, +1 more.
update_secretAIdempotentInspect
Rotate an existing workspace secret. Fails if the name is missing. The value is never returned. Use when replacing a credential that list_secrets already shows.
Inputs: name (required), value (required), description.
update_skillBInspect
Update an existing custom skill (name, description, body, or keywords). Use when refining agent guidance.
Inputs: id (required), body, name, keywords, +1 more.
update_toolAInspect
Update an existing custom HTTP tool. Omitted fields are preserved. Use when changing URL, method, headers, or schemas. Auth values should use "{{SECRET_NAME}}". Registered vendor URLs belong in metaphor.integration, not execute.url.
Inputs: id (required), name, execute, description, +2 more.
update_triggerAIdempotentInspect
Change the name or any part of a trigger's working copy (app, event, scope, fetch code); what is left out stays as it is. Nothing that runs changes until it is published, and the last test no longer counts: test again first. A published trigger can't switch to another app or event; create a new one.
Inputs: triggerId (required), name, event, fetch, +2 more.
upsert_data_table_recordAIdempotentInspect
Full replace of record data by recordKey (creates if missing). Last-write-wins with no version check. Use for idempotent imports; prefer update_data_table_record when concurrent writers must not overwrite each other.
Inputs: data (required), recordKey (required), dataTableId (required).
verify_playbookARead-onlyIdempotentInspect
Audit a playbook's dataflow and configuration without running it. Checks input mappings, step output references, sandbox network/secret usage, and metaphor.tool/agent/dataTable ids. Use after create_playbook or update_playbook, or before run_playbook, when diagnosing wiring issues in automations such as weekly reports, lead enrichment, invoice approval, or ticket triage.
Inputs: id (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
CRM, inbox, meetings, revenue intelligence, analytics and automations for AI agents.
Run business workflows and route requests for human approval from AI agents.
Hosted AI agents for revenue teams: your CRM and tools, plus Apollo, FullEnrich and more at cost.
Agentic-AI operations platform for any business — automate sales, finance, support and back-office w
Related MCP Servers
FlicenseNot gradedqualityCmaintenanceEnables AI clients to inspect and safely modify Salesforce and HubSpot data through read-only and permission-gated tools for deduplication, normalization, enrichment, assignment, job-change detection, record creation, mass updates, and segmentation, with auditable and reversible writes.1-- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to run an entire B2B prospecting workflow from one request: searching people and companies, enriching verified emails and phone numbers, monitoring buying signals, researching accounts, and building lists. It also generates and runs multi-step email, LinkedIn and calling campaigns, with everything created as drafts until explicitly launched.MIT
- AlicenseNot gradedqualityBmaintenanceEnables building and running visual workflows conversationally inside Claude, orchestrating your existing connector tools step-by-step without storing credentials or owning integrations.11 npmMIT

PumpGTM MCP serverofficial
AlicenseNot gradedqualityBmaintenanceEnables AI agents to find intent-signaling buyers and run LinkedIn, email, and X outreach from connected accounts while respecting platform limits. Every reply is handed back to a human for approval or decision.3MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.