SpecStream
Server Details
The live web data engine for teams and agents: Search API, Extract API and Deep Search.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- specstream/specstream-tools
- GitHub Stars
- 0
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 165 tools
Descriptions are unusually careful, using explicit "Fits when... Not for..." guidance and disambiguating near-neighbours (compile vs search_web vs read_pages, prepare_reconciliation vs apply_reconciliation). But with 165 tools there are many closely overlapping families an agent can easily misselect: get_find vs get_delivery vs list_find_rows vs get_capacity_execution, cancel_find vs cancel_research vs cancel_investigation_batch, share_find vs share_collection vs create_spec_link, and the browser_action catch-all vs the 30+ browser_* steps.
The bulk of names follow a predictable snake_case verb_noun pattern (get_*, list_*, create_*, update_*, prepare_*, save_*, revoke_*), and the browser_* family is consistently prefixed. Minor deviations exist: noun-first or verbless names such as spec_options, research_price, compile, teach_web_source, reconcile_find_commerce, and a capacity_* namespace overlapping with get_*/create_* forms.
165 tools is an extreme surface for a single server, far beyond the 3-15 sweet spot even for a broad domain. Multiple large sub-systems (Playwright browser stepping, billing/quote lifecycle, reconciliation, lists, projects, sharing, web sources, research rules) are flattened into one undifferentiated list, making discovery and correct selection costly.
Coverage across the inferred domain is genuinely broad: quote/approve/authorize/start/cancel/refund flows, list lifecycle, imports, destinations, sharing, projects/members, API keys and audit are all present. Minor gaps remain (e.g. no explicit delete for lists or saved requests beyond archive, limited note-update operations), but core workflows appear closed.
Publisher details
- Operator
- SpecStream · Publisher source
- Operator website
- https://specstream.ai/?utm_source=glama · Publisher source
- Vendor relationship
- First-party · Publisher source
- Documentation
- https://specstream.ai/docs/mcp · Publisher source
- Privacy policy
- Not provided
- Trust center
- Not available
- Restrictions
- Sign in with a SpecStream account through OAuth. Each run shows its price first. · Publisher source
Available Tools
165 toolsadd_noteLeave a noteADestructiveInspect
Adds a note to a project or a record. Each person the note @mentions gets an email.
Inputs: id (required), text (required), expected_version (required), record_id, +2 more. Returns: notes, version
apply_reconciliationFill a list from chosen pairsAInspect
Fills empty cells into a new list from the counterpart rows of the pairs the user chose. It applies only pairs the user chose from the proposed ones. It takes the exact comparison hash. Original cells and lists stay unchanged, and unmatched rows stay in the new list. Filled cells are inputs, not checked results. Research on them needs a normal quote and approval.
Inputs: name (required), choices (required), import_id (required), comparison_id (required), +1 more. Returns: id, name, rows, imported, +8 more
approve_investigation_batchApprove a run over list rowsCDestructiveInspect
Approves the quote's exact maximum. The searches then run as ordinary finds, within that maximum.
Inputs: batch_id (required), quote_hash (required), confirm_max_cents (required). Returns: name, rows, state, works, +10 more
approve_research_ruleApprove automatic researchADestructiveInspect
Approves the rule's total spending limit. Rows that qualify later then run with no new approval, until the rule reaches the limit or the expiry.
Inputs: rule_id (required), review_hash (required), confirm_total_cents (required). Returns: name, rows, runs, state, +21 more
archive_specArchive or restore a requestCDestructiveInspect
Archives a saved request, or restores it. Its results stay.
Inputs: spec_id (required), expected_revision_id (required), archived. Returns: spec_id, archived, revision, account_id, +4 more
authorize_capacity_workApprove a quoteADestructiveInspect
Fits when the user has seen a quote and said yes to its maximum. Not for getting a price. Starts the work on that quote. It takes the quote_id and quote_hash, the ceiling the user approved in cents as confirm_max_cents, and a stable idempotency_key. SpecStream then reads the live web and charges for delivered results at the quoted rate, plus the checks the quote lists, never more than the ceiling.
Inputs: quote_id (required), quote_hash (required), idempotency_key (required), confirm_max_cents (required), +1 more. Returns: state, replay, find_id, maximum_cents, +1 more
browser_actionRun a browser step by nameADestructiveInspect
Runs one step in an open browser by a Playwright MCP tool name and its arguments. The browser_ tools are the same steps with their own arguments; use this one only for a step they do not list. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: tool (required), session_id (required), arguments. Returns: url, error, title, result, +3 more
browser_clickClickBDestructiveInspect
Perform click on a web page. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: target (required), session_id (required), button, element, +2 more. Returns: The page after the step, and what the step cost.
browser_console_messagesGet console messagesCDestructiveInspect
Returns all console messages. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: level (required), session_id (required), all, filename. Returns: The page after the step, and what the step cost.
browser_dragDrag mouseBDestructiveInspect
Perform drag and drop between two elements. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: endTarget (required), session_id (required), startTarget (required), endElement, +1 more. Returns: The page after the step, and what the step cost.
browser_dropDrop files or data onto an elementADestructiveInspect
Drop files or MIME-typed data onto an element, as if dragged from outside the page. At least one of "paths" or "data" must be provided. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: target (required), session_id (required), data, paths, +1 more. Returns: The page after the step, and what the step cost.
browser_emulate_mediaEmulate media featuresADestructiveInspect
Emulate CSS media features for the page, for example switch between the light and dark color scheme. Omitted parameters are left unchanged; null clears an override. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: session_id (required), media, contrast, colorScheme, +2 more. Returns: The page after the step, and what the step cost.
browser_evaluateEvaluate JavaScriptBDestructiveInspect
Evaluate JavaScript expression on page or element. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: function (required), session_id (required), target, element, +1 more. Returns: The page after the step, and what the step cost.
browser_file_uploadUpload filesBDestructiveInspect
Upload one or multiple files. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: session_id (required), paths. Returns: The page after the step, and what the step cost.
browser_fill_formFill formBDestructiveInspect
Fill multiple form fields. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: fields (required), session_id (required). Returns: The page after the step, and what the step cost.
browser_findFind in page snapshotADestructiveInspect
Search the accessibility snapshot of the current page for text or a regular expression. Returns matching snapshot nodes with a few lines of surrounding context (like search snippets), each shown under its path from the root of the tree, which is cheaper than capturing the whole snapshot when you only need to locate an element and its ref. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: session_id (required), text, regex, filename, +1 more. Returns: The page after the step, and what the step cost.
browser_handle_dialogHandle a dialogCDestructiveInspect
Handle a dialog. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: accept (required), session_id (required), promptText. Returns: The page after the step, and what the step cost.
browser_hoverHover mouseADestructiveInspect
Hover over element on page. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: target (required), session_id (required), element. Returns: The page after the step, and what the step cost.
browser_network_requestShow network request detailsADestructiveInspect
Returns full details (headers and body) of a single network request, or a single part if part is set. Use the number from browser_network_requests. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: index (required), session_id (required), part, filename. Returns: The page after the step, and what the step cost.
browser_network_requestsList network requestsADestructiveInspect
Returns a numbered list of network requests since loading the page. Use browser_network_request with the number to get full details. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: static (required), session_id (required), all, filter, +1 more. Returns: The page after the step, and what the step cost.
browser_press_keyPress a keyBDestructiveInspect
Press a key on the keyboard. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: key (required), session_id (required). Returns: The page after the step, and what the step cost.
browser_readRead the pageBDestructiveInspect
The current page's text as Markdown (or plain text), read from the live page by the same reader as the Extract API, with the page's URL and the time of the read. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: session_id (required), format. Returns: The page after the step, and what the step cost.
browser_resizeResize browser windowADestructiveInspect
Resize the browser window. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: width (required), height (required), session_id (required). Returns: The page after the step, and what the step cost.
browser_select_optionSelect optionBDestructiveInspect
Select an option in a dropdown. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: target (required), values (required), session_id (required), element. Returns: The page after the step, and what the step cost.
browser_snapshotPage snapshotBDestructiveInspect
Capture accessibility snapshot of the current page, this is better than screenshot. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: session_id (required), boxes, depth, target, +1 more. Returns: The page after the step, and what the step cost.
browser_tabsManage tabsADestructiveInspect
List, create, close, or select a browser tab. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: action (required), session_id (required), url, index, +1 more. Returns: The page after the step, and what the step cost.
browser_take_screenshotTake a screenshotADestructiveInspect
Take a screenshot of the current page. Use browser_snapshot for actions by default. With the vision capability enabled, use screenshot coordinates with the coordinate-based mouse tools for elements not exposed in the accessibility snapshot. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: scale (required), session_id (required), type, target, +3 more. Returns: The page after the step, and what the step cost.
browser_typeType textBDestructiveInspect
Type text into editable element. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: text (required), target (required), session_id (required), slowly, +2 more. Returns: The page after the step, and what the step cost.
browser_wait_forWait forADestructiveInspect
Wait for text to appear or disappear or a specified time to pass. When both text and textGone are provided, waits for the first one to happen. Takes the session_id from open_browser. 0.1¢ a step that runs.
Inputs: session_id (required), text, time, textGone. Returns: The page after the step, and what the step cost.
cancel_findStop a runADestructiveInspect
Asks a run in progress to stop. The run keeps what it already found.
Inputs: find_id (required). Returns: status, find_id, cancel_requested
cancel_investigation_batchStop a run over list rowsADestructiveInspect
Stops searches that did not start, and cancels running ones where possible. Finished results stay with the user.
Inputs: batch_id (required). Returns: name, rows, state, works, +10 more
cancel_researchStop a research reportADestructiveInspect
Stops a research run at its next step. A cancelled run is not charged and writes no report. Returns the run.
Inputs: id (required). Returns: id, depth, error, input, +9 more
claim_free_capacityClaim the free startAInspect
Claims the one-time free balance for an eligible, verified account.
Inputs: no named parameters. Returns: low, lots, currency, auto_topup, +15 more
close_browserClose a browserADestructiveInspect
Closes a browser session and settles its charge for the seconds it ran and its steps. Call it when the task is done: an open session is charged until it times out. Files the page downloaded stay 10 minutes. A repeat call answers the same.
Inputs: session_id (required). Returns: id, bytes, price, limits, +20 more
combine_collectionsCombine listsAInspect
Merges collections the caller owns into one, and keeps each record once with its sources.
Inputs: name (required), request_id (required), collection_ids (required), project_id. Returns: id, ast, kind, name, +22 more
compileUnderstand a requestARead-onlyIdempotentInspect
Fits when the user wants a list or table from the live web, every match or as many as asked, with the same facts on each row, also when a web search returned only part of the list: companies and local businesses, products and stores, jobs, papers, clinical trials, grants, tenders, datasets, repositories and packages, apps, books or podcasts. Not for one fact or a quick answer, which the fast web search gives, or for the text of known pages. A list the user already has is imported first, and its quote takes this reading. Reads the request in everyday words in any language. Returns how SpecStream understood it: what to find, the conditions for every result, the facts to bring back (required or optional), how many, and in what order. The returned ast_hash identifies this reading. For a list the user has, parent_find_id and the operation add facts to it, extend it or refresh it. In that case, what it finds and its conditions stay the same, and only the new facts or the new count are read. The caller needs read access to that list. Nothing runs and nothing is charged.
Inputs: q (required), locale, retained, named_phrases. Returns: ast, chips, label, notice, +11 more
create_api_keyCreate an API keyBDestructiveInspect
Creates an API key with permissions, an expiry date and a spending limit. A new key needs the account owner's yes.
Inputs: name, scopes, project_id, engine_rule, +4 more. Returns: name, token, key_id, prefix, +8 more
create_capacity_refund_claimReport a wrong resultBInspect
Sends charged results that do not meet the request's conditions for review and a refund.
Inputs: reason (required), unit_ids (required), reservation_id (required), idempotency_key (required). Returns: object
create_capacity_scheduleCreate a monitorCDestructiveInspect
Runs a request on a schedule, hourly to yearly, within a limit per run and per period. Each run reads the live web. A refresh bills half the rate for a changed row and 1 cent for an unchanged row. monthly_maximum_cents is the most the monitor can spend in one UTC calendar month. It equals the limit per period, and that limit covers both the scheduled runs and the runs started on demand. scheduled_monthly_cents is the most the scheduled runs alone can spend in their busiest UTC calendar month: the number of runs in that month times the limit per run, never more than the limit per period.
Inputs: scope (required), valid_until (required), first_due_at (required), period_cents (required), +4 more. Returns: object
create_destination_syncKeep a tool up to dateCDestructiveInspect
Sends new and changed records to one destination on each change, with the same mapping.
Inputs: collectionId (required), connection_id (required), destination_sha256 (required), keepSending. Returns: result
create_investigation_functionMake a request run per rowAInspect
Makes a saved request run once per row, with some of its values as typed inputs. Nothing runs and nothing is charged.
Inputs: name (required), spec_id (required), parameters (required), revision_id (required). Returns: name, request, spec_id, created_at, +7 more
create_projectCreate a projectBInspect
Creates a project with its own spending limit.
Inputs: name (required), monthly_spend_cap_cents. Returns: name, role, version, can_spend, +4 more
create_specSave a requestBInspect
Saves a reviewed request as its first version, so it can be rerun, shared or changed.
Inputs: name (required), q, ast, context, +2 more. Returns: spec_id, archived, revision, account_id, +4 more
create_spec_linkShare a request as a linkADestructiveInspect
Creates a link to the wording and conditions of a saved request. The link never shares results or the balance.
Inputs: spec_id (required), revision_id (required), expected_ast_hash (required), expires_seconds, +1 more. Returns: link_id, expires_at, public_url, snapshot_hash
create_webhookAdd a webhookADestructiveInspect
Adds a webhook that sends a signed event to a URL when a run finishes, a monitor finds something new, or records are added. The response shows the signing secret once.
Inputs: url (required), events, event_name. Returns: events, secret, hook_id, event_name
delegate_accessDelegate a budgetCDestructiveInspect
Lets a key or project run specific work within a per-run and monthly limit.
Inputs: actor_key (required), operations (required), valid_until (required), period_cents (required), +4 more. Returns: object
edit_specChange what to find or bring backARead-onlyIdempotentInspect
Changes a condition or a fact to bring back without rewording the request. It takes the revised reading as ast, with the edit operations. Returns a checked reading with a new ast_hash for the next quote. Results that already ran stay as they were.
Inputs: ast (required), operations (required). Returns: ast, chips, label, notice, +11 more
export_findDownload the resultsAInspect
Creates a download link for a run's results as CSV, Excel, JSON or HTML, or as KML and GeoJSON for records with places. The link works for 24 hours.
Inputs: format (required), find_id (required). Returns: filename, single_use, download_url, expires_in_s
filter_by_areaFilter by map areaCInspect
Narrows a collection to the records inside an area on the map.
Inputs: collection_id (required), bbox, record_ids, source_collection_ids. Returns: id, ast, kind, name, +22 more
filter_collectionNarrow a listCInspect
Keeps the selected records of a collection, by ID, by source or by location.
Inputs: collection_id (required), bbox, record_ids, source_collection_ids. Returns: id, ast, kind, name, +22 more
get_accountRead the accountBRead-onlyIdempotentInspect
Reads the account and its payment history.
Inputs: no named parameters. Returns: email, finds, account_id
get_browserRead a browser sessionARead-onlyIdempotentInspect
Reads a browser session: running or ended, why it ended, its seconds, steps, bytes and charge so far. No charge.
Inputs: session_id (required). Returns: id, bytes, price, limits, +20 more
get_capacity_balanceRead the balanceARead-onlyIdempotentInspect
Reads the account's balance in cents: what is available, what is held for work in progress, and the monthly limit and what it has used. A balance is not permission to spend. Paid work starts only from a quote the user approved.
Inputs: project_id. Returns: low, lots, currency, auto_topup, +15 more
get_capacity_catalogRead pricesARead-onlyIdempotentInspect
Reads the plans, per-result rates and top-up options.
Inputs: no named parameters. Returns: object
get_capacity_entitlementsCheck what the plan includesARead-onlyIdempotentInspect
Reads the account's plan, what it includes, such as schedules, CRM sync and single sign-on, and how many members it has. A feature outside the plan returns plan_required with the plan that includes it.
Inputs: project_id. Returns: plan, unlock, plan_name, capabilities, +4 more
get_capacity_executionRead what a run costCRead-onlyIdempotentInspect
Reads the finished work, what it cost, and the caller's access to its results.
Inputs: find_id (required). Returns: find_id, currency, shortfall, next_action, +18 more
get_capacity_quoteRead a quoteARead-onlyIdempotentInspect
Reads one quote by its quote_id: the most the work can cost in cents, what it covers, its terms, when it expires and whether it can still be approved, plus the run already started from it, if any. Nothing runs and nothing is spent.
Inputs: quote_id (required). Returns: scope, terms, currency, quote_id, +10 more
get_capacity_refund_claimCheck a reportCRead-onlyIdempotentInspect
Reads the status of a reported wrong result.
Inputs: claim_id (required). Returns: object
get_capacity_requestFind work already startedARead-onlyIdempotentInspect
Reads the work started with an idempotency_key, so a retry that is not sure the first call worked does not start the work twice.
Inputs: request_key (required), project_id. Returns: object
get_collectionRead a listBRead-onlyIdempotentInspect
Reads one list: its records, columns, record count and duplicates, the run or import it came from, and the team fields on it.
Inputs: collection_id (required). Returns: id, ast, kind, name, +22 more
get_collection_changesSee what changedBRead-onlyIdempotentInspect
Compares a list with its earlier results: each changed fact, before and after, and where it was seen.
Inputs: collection_id (required). Returns: runId, checks, status, changes, +7 more
get_connection_columnsRead a destination's fieldsBRead-onlyIdempotentInspect
Reads the columns or properties a destination has, in the tool's own words. Nothing is sent.
Inputs: selection (required), connection_id (required). Returns: result
get_deliveryOpen delivered resultsARead-onlyIdempotentInspect
Reads a run's finished results with the request version and the sources behind them.
Inputs: find_id (required). Returns: rows, find_id, lineage, ast_hash, +4 more
get_findCheck on a runARead-onlyIdempotentInspect
Fits when the user asks how a run is going or what it has cost so far. Not for reading its rows. Reads a run's request, its progress and its cost so far.
Inputs: find_id (required). Returns: spec, status, find_id, outcome, +21 more
get_find_commerceCheck a run's paymentBRead-onlyIdempotentInspect
Reads what a run was charged, what it delivered, and who can open it.
Inputs: find_id (required). Returns: find_id, currency, shortfall, next_action, +18 more
get_find_decisionsSee why each result qualifiedBRead-onlyIdempotentInspect
Reads each condition check on a run's results: yes, no or unsure, the value seen, and its source.
Inputs: find_id (required), limit, cursor. Returns: find_id, decisions, next_cursor, execution_manifest
get_find_lineageSee where a run came fromBRead-onlyIdempotentInspect
Reads which version of a request a run used, and the earlier run it continued.
Inputs: find_id (required). Returns: find_id, spec_id, operation, revision_id, +1 more
get_input_researchReview imported researchARead-onlyIdempotentInspect
Reads each original row of an imported list, the identity decision with its source, any unresolved namesakes and the delivered record. The name and location in the input are search hints, not verified facts. An identity without evidence stays missing and is not a Match. No charge and no new research.
Inputs: find_id (required). Returns: rows, provenance, source_name, input_access, +1 more
get_investigation_batchRead a run over list rowsCRead-onlyIdempotentInspect
Reads each input row, its search status and the list it led to.
Inputs: batch_id (required). Returns: name, rows, state, works, +10 more
get_investigation_functionRead a request that runs per rowCRead-onlyIdempotentInspect
Reads a saved request that runs per row, and its typed inputs.
Inputs: function_id (required). Returns: name, request, spec_id, created_at, +7 more
get_investigation_inputsList a list's columns for inputsARead-onlyIdempotentInspect
Lists the columns of a list with example values, for the choice of which column feeds each input.
Inputs: collection_id (required). Returns: fields, project_id, total_rows, collection_id
get_investigation_parametersList a request's possible inputsBRead-onlyIdempotentInspect
Lists the values in a saved request that can become inputs.
Inputs: spec_id (required), revision_id. Returns: spec_id, parameters, revision_id
get_investigation_resultsRead a run's results by rowARead-onlyIdempotentInspect
Reads each input row with the results it led to, and their sources. Input values are never sources for results.
Inputs: batch_id (required). Returns: inputs, schema, batch_id, deliveries, +1 more
get_meCheck spending limitsARead-onlyIdempotentInspect
Reads the spending limit for this connection and how much of it is left.
Inputs: no named parameters. Returns: email, guest, account_id, spend_ceiling_set, +5 more
get_projectRead a projectARead-onlyIdempotentInspect
Reads one project: its name, its owner, the caller's role in it, whether the caller can spend in it, its monthly spending limit in cents and its version.
Inputs: project_id (required). Returns: name, role, version, can_spend, +4 more
get_reconciled_inputRead where filled cells came fromARead-onlyIdempotentInspect
Reads a list filled from a comparison: each row, with the chosen pair and the other list's record that filled each cell. Filled cells are inputs, not checked results.
Inputs: import_id (required). Returns: name, rows, import_id, comparison_id, +1 more
get_reconciliationRead a comparisonARead-onlyIdempotentInspect
Reads one comparison of two lists: the proposed pairs, the alternatives for each row, the conflicts and the missing values, with counts. Returns an error if either list changed after the comparison was prepared.
Inputs: comparison_id (required). Returns: name, rows, metric, scopes, +8 more
get_reconciliation_fieldsList fields to match onARead-onlyIdempotentInspect
Reads the fields of a list the user can open, with example values and the record count, to choose the fields that match it against another list. Changes nothing.
Inputs: collection_id (required). Returns: name, fields, records, collection_id
get_relationshipsSee what records shareARead-onlyIdempotentInspect
Groups records by the values they share in a column.
Inputs: collection_id (required), field_id. Returns: label, groups, missing, version, +2 more
get_researchRead a research reportARead-onlyIdempotentInspect
Reads a research run: its status, its stage (planning, searching, reading, writing, checking or done), how many searches, pages and sources it has, and the report once the run is completed. The report is markdown with numbered citations, or JSON for a JSON Schema with the citations of each field. Each citation names the page, its title, when it was read, the passage and whether that passage supports the sentence. A call with a progress token gets progress notifications while the run works, for up to 45 seconds. Nothing is charged.
Inputs: id (required). Returns: id, depth, error, input, +9 more
get_research_ruleRead an automatic research ruleBRead-onlyIdempotentInspect
Reads a rule's limits, what it spent, which rows are waiting, and its recent runs.
Inputs: rule_id (required), offset. Returns: name, rows, runs, state, +21 more
get_research_rule_resultsRead automatic research resultsARead-onlyIdempotentInspect
Reads every run of the rule with its input rows and results.
Inputs: rule_id (required). Returns: runs, rule_id, current_inputs, original_deliveries_unchanged, +1 more
get_revisitRead what a recheck foundARead-onlyIdempotentInspect
Reads one page of the records a recheck looked at: what still holds, what changed, and what it could not reach. next_after is the after value for the next page, and it is null on the last page.
Inputs: find_id (required), after, limit. Returns: find_id, summary, outcomes, next_after
get_specRead a saved requestBRead-onlyIdempotentInspect
Reads a saved request: its project, its current version number and id, whether it is archived, and when it was created and last changed.
Inputs: spec_id (required). Returns: spec_id, archived, revision, account_id, +4 more
get_spec_revisionRead one version of a saved requestBRead-onlyIdempotentInspect
Reads one version of a saved request as it was saved: its reading of the request with the ast_hash, its name and notes, and the version it came from.
Inputs: spec_id (required), revision_id (required). Returns: ast, name, label, context, +6 more
get_web_searchRead a past web searchBRead-onlyIdempotentInspect
Opens one past fast search by its id, from the account's search history or from the answer of the search itself, exactly as it was given: the answer, the rows with the page of each value, the search results read and, for a list, the hand-off to the full table. Nothing is charged.
Inputs: answer_id (required). Returns: id, kind, news, page, +20 more
get_web_sourceRead a website sourceBRead-onlyIdempotentInspect
Reads one saved website source: its page address, its column headers, and how each field is read from the page and what happens when the page changes.
Inputs: source_id (required). Returns: name, reuse, fields, headers, +5 more
get_web_source_previewRead a source previewARead-onlyIdempotentInspect
Reads one preview of a website source: its rows, where each cell sits on the page, the missing cells, and how each field is read.
Inputs: preview_id (required). Returns: name, rows, count, reuse, +12 more
import_csvImport a CSV or Excel fileAInspect
Fits when the user already has a list as a CSV or Excel file and wants details filled in or each row checked. Not for finding new rows on the web, or for a linked Google Sheet or Airtable view. Imports the file with its names, websites, places and any other columns. Nothing is researched or charged. Returns the import id, which a quote takes as source_import_id to fill in details or check each row.
Inputs: name (required), format (required), import_id (required), sheet, +4 more. Returns: id, name, rows, imported, +8 more
import_from_sourceImport from a connected sheetBDestructiveInspect
Imports a Google Sheet, Excel workbook or Airtable view. It can also check the source every day for new and changed rows.
Inputs: link (required), name (required), import_id (required), connection_id (required), +6 more. Returns: result
invite_project_memberInvite someone to a projectBDestructiveInspect
Emails a project invitation with a role, a spending limit and an expiry date.
Inputs: email (required), project_id (required), request_id (required), role, +2 more. Returns: id, role, email, status, +1 more
list_activityRead activityARead-onlyIdempotentInspect
Reads what happened across the account and its projects, newest first: each event, who did it and when. next_cursor is null on the last page.
Inputs: limit, cursor, project_id. Returns: items, next_cursor
list_audit_eventsRead the audit logARead-onlyIdempotentInspect
Reads sign-ins, new keys, approved spending, exports, monitors and imports from the last 12 months.
Inputs: kind, limit, cursor, project_id, +1 more. Returns: events, next_cursor, retention_months
list_capacity_occurrencesSee each monitor runARead-onlyIdempotentInspect
Reads a monitor's completed, blocked and missed runs, with the reason for each.
Inputs: schedule_id (required), limit, before. Returns: object
list_capacity_schedulesList monitorsARead-onlyIdempotentInspect
Lists the account's monitors and whether each is running or paused. monthly_maximum_cents is the most a monitor can spend in one UTC calendar month, equal to its limit per period. scheduled_monthly_cents is the most its scheduled runs alone can spend in their busiest UTC calendar month.
Inputs: after, limit, project_id. Returns: object
list_cloud_importsList linked sheetsBRead-onlyIdempotentInspect
Lists the linked sheets and tables, and when each was last checked.
Inputs: no named parameters. Returns: result
list_collection_investigationsList runs over a listBRead-onlyIdempotentInspect
Lists the recent runs started from a list.
Inputs: collection_id (required). Returns: batches
list_collectionsList collectionsARead-onlyIdempotentInspect
Fits when the user asks which lists they have. Not for finding one by name or text, which the workspace search does. Lists the imported lists and results the caller can open.
Inputs: sort, after, limit, project_id. Returns: next_after, collections
list_connected_toolsSee tools you can connectARead-onlyIdempotentInspect
Lists the sheets, databases, CRMs, automation and chat tools SpecStream can send to.
Inputs: no named parameters. Returns: result
list_connection_destinationsPick where records goBRead-onlyIdempotentInspect
Lists the sheets, tables, bases and CRM objects in a connected account.
Inputs: connection_id (required), cursor, base_id. Returns: result
list_connectionsList connected toolsARead-onlyIdempotentInspect
Lists the spreadsheets, databases, CRMs and automation tools this account connected.
Inputs: no named parameters. Returns: result
list_destination_exportsRead what was sentBRead-onlyIdempotentInspect
Reads each send: where it went, how many records arrived, and which were updated or skipped.
Inputs: no named parameters. Returns: result
list_destination_syncsList continuous sendsBRead-onlyIdempotentInspect
Lists every destination that receives new records, and whether each is on.
Inputs: no named parameters. Returns: result
list_find_rowsRead a run's resultsARead-onlyIdempotentInspect
Fits when a run has delivered results and the user wants the rows or the table. Not for a run's progress or cost. Reads the delivered results, one page at a time, with the source behind each condition a page states. A condition no page states yet is listed on its result. next_cursor is null on the last page. Every delivered result is charged. The markdown and csv formats return the asked columns in order as one table, with a source link for each value.
Inputs: find_id (required), limit, cursor, format, +1 more. Returns: rows, table, format, status, +10 more
list_investigation_functionsList requests that run per rowARead-onlyIdempotentInspect
Lists the saved requests in a project that can run once per row.
Inputs: project_id. Returns: functions
list_notesRead notesCRead-onlyIdempotentInspect
Reads the notes on a project or a record.
Inputs: record_id, project_id, collection_id. Returns: notes, version
list_notificationsRead notificationsARead-onlyIdempotentInspect
Reads mentions, finished work and blocked monitors.
Inputs: tab, limit, cursor. Returns: next_cursor, notifications
list_project_membersList project membersBRead-onlyIdempotentInspect
Lists the people in a project and their roles.
Inputs: project_id (required), limit, cursor. Returns: members, next_cursor
list_projectsList projectsARead-onlyIdempotentInspect
Lists the projects the caller can open, one page at a time, with the caller's role in each. next_cursor is null on the last page.
Inputs: limit, cursor. Returns: projects, next_cursor
list_reconciliationsList recent comparisonsARead-onlyIdempotentInspect
Lists the 20 newest comparisons that include a list, each with its id, name, date and summary counts. SpecStream keeps a comparison for 30 days. A list filled from a comparison stays as its own list.
Inputs: collection_id (required). Returns: comparisons
list_request_templatesList example requestsARead-onlyIdempotentInspect
Fits when the user asks what they can request, or wants an example request to start from. Not for a request the user has already written, which is read as it is. Lists example requests that work for each kind of record, each with its wording and kind. Nothing runs from an example until a quote for it is approved.
Inputs: record_type. Returns: version, templates
list_researchList research reportsARead-onlyIdempotentInspect
Lists the account's research runs, newest first, each with its status, depth and progress; next_after continues the list. Nothing is changed or charged.
Inputs: after, limit. Returns: data, has_more, next_after
list_research_rulesList automatic research rulesCRead-onlyIdempotentInspect
Lists the automatic research rules on a list.
Inputs: collection_id (required). Returns: rules
list_spec_findsSee every run of a requestBRead-onlyIdempotentInspect
Lists the runs of each version of a saved request.
Inputs: spec_id (required), limit, cursor. Returns: finds, next_cursor
list_spec_revisionsRead a request's historyARead-onlyIdempotentInspect
Lists every saved version of a request.
Inputs: spec_id (required), limit, cursor. Returns: revisions, next_cursor
list_specsList saved requestsARead-onlyIdempotentInspect
Lists the saved requests the caller can open, one page at a time, with the id and current version of each. next_cursor is null on the last page.
Inputs: limit, cursor, archived, project_id. Returns: specs, next_cursor
list_starredList starred workARead-onlyIdempotentInspect
Lists the starred requests, collections and projects.
Inputs: no named parameters. Returns: starred, version
list_viewsList saved viewsCRead-onlyIdempotentInspect
Lists the views saved on a collection.
Inputs: collection_id (required). Returns: views, version, collection_id
list_web_searchesList your web searchesARead-onlyIdempotentInspect
Lists the account's past fast searches, newest first: each question with when it was asked, its kind (fact or list), the answer and the page it was read from, the number of rows, and the search results its values come from (sources). q keeps the searches whose words contain it. An answer the account already paid for is read here again without a new search. next_cursor leads to older searches. Nothing is charged.
Inputs: q, limit, cursor. Returns: results, next_cursor
list_web_sourcesList website sourcesARead-onlyIdempotentInspect
Lists the website sources saved in a project, each with its id, name and page address.
Inputs: no named parameters. Returns: sources
list_your_fieldsRead a list's team fieldsBRead-onlyIdempotentInspect
Reads the status, checkbox and text fields a team added to a list, with the value of each field on each record and the fields' version.
Inputs: collection_id (required). Returns: fields, values, version, collection_id
open_browserOpen a browserADestructiveInspect
Opens a real Chromium browser on SpecStream's server and returns its session id. Drive it with the browser_ tools (browser_navigate, browser_snapshot, browser_click, browser_type, browser_take_screenshot, browser_read and the others) with that session_id, or hand its connect_url to Playwright, Puppeteer, browser-use or Stagehand. Use it for pages that need a real browser: forms, clicks, pages drawn by scripts. For the text of known pages read_pages is faster and cheaper. A page whose site opted out, whose robots.txt disallows SpecStream or that is outside allowed_domains answers blocked. The session ends at close_browser, after idle_timeout_s with no step or connection, or at timeout_s (default 300 seconds, at most 3,600). Charged $0.10 an hour by the second while it runs (at most 10¢ a session), plus $1 per 1,000 steps (0.1¢ a step). A session that fails to start is not charged.
Inputs: headless, metadata, viewport, timeout_s, +4 more. Returns: id, bytes, price, limits, +20 more
patch_research_rulePause, resume or delete a ruleADestructiveInspect
Pauses, resumes or deletes a rule. A pause holds queued searches, and running searches finish. A resume never raises the limits.
Inputs: state (required), rule_id (required), expected_version (required). Returns: name, rows, runs, state, +21 more
prepare_investigation_batchQuote a run over list rowsAInspect
Binds the selected rows to the inputs and returns one quote: the distinct searches, rows with missing inputs, searches reused from earlier work, and one maximum. Nothing runs yet.
Inputs: bindings (required), record_ids (required), request_id (required), function_id (required), +2 more. Returns: name, rows, state, works, +10 more
prepare_reconciliationCompare two listsAInspect
Proposes which rows in one list correspond to rows in another list. It takes matching groups of exact or close-spelling rules, and conflict fields that rule a pair out. A pair matches when every rule in one group agrees. Returns the proposed pairs, alternatives, conflicts and missing values. A close spelling is not proof of the same company, and a single candidate is not a chosen pair. No search and no charge.
Inputs: fill (required), left (required), name (required), right (required), +6 more. Returns: name, rows, metric, scopes, +8 more
prepare_research_rulePrepare automatic researchAInspect
Prepares a rule that researches rows when they first become ready or when their inputs change, with a per-run and a total spending limit. Nothing runs until the rule is approved.
Inputs: name (required), bindings (required), request_id (required), function_id (required), +9 more. Returns: name, rows, runs, state, +21 more
preview_cloud_listPreview a connected sheetARead-onlyIdempotentInspect
Reads the first rows and columns of a Google Sheet, Excel workbook or Airtable view before an import.
Inputs: link (required), connection_id (required), sheet_id. Returns: result
preview_web_sourceRead a website source againARead-onlyIdempotentInspect
Reads a saved source's page again and returns a new preview of its rows, with the count and the missing cells. Earlier lists stay unchanged.
Inputs: source_id (required). Returns: name, rows, count, reuse, +12 more
put_project_memberSet a member's role and limitBDestructiveInspect
Sets a person's role in a project, editor or viewer, and their own monthly spending limit.
Inputs: role (required), account_id (required), project_id (required), monthly_spend_cap_cents. Returns: role, account_id, monthly_spend_cap_cents
quote_capacity_workGet a quoteAInspect
Fits when the user wants the price of a request before anything runs. Not for starting the work, which waits for the user's yes to the quote. Prices the reading identified by ast_hash. Operation find prices a new request. Operation extend prices more results, add_facts prices more facts on results the user has, and refresh prices a recheck of them. These three take parent_find_id. For an imported list, operation find with source_import_id and its source_row_ids fills in details or checks each row. The quote shows the most the work can cost. Nothing is spent.
Inputs: ast_hash (required), operation, project_id, revision_id, +6 more. Returns: scope, terms, currency, quote_id, +10 more
read_pagesRead pagesAInspect
Fits when the user gives URLs, or wants the full text of pages a search returned. Not for finding pages or answering a question, which the fast web search does, or for building a list. Reads up to 20 web pages or documents (PDF, Word, Excel) live and returns each one's main text, as Markdown (format markdown, the default) or plain text (format text), with its title and the date it says it was published. Each result has status: ok, blocked, not_found, empty, unsupported, invalid_url or failed, and read_from: live, or site or archive when the page was slow or refused a live read and its site's feed, SpecStream's own dated copy or an archive capture gave the text (read_from_url names the feed, captured_at says when the copy or capture was taken). A page is charged only when its status is ok, $0.0008 a page from the account's balance, with no quote or approval.
Inputs: urls, pages, format, include. Returns: results, took_ms, price_usd, request_id
reconcile_destination_exportCheck an uncertain sendARead-onlyIdempotentInspect
Checks what arrived in a destination, so nothing is sent twice.
Inputs: export_id (required), connection_id (required). Returns: result
reconcile_find_commerceRecover a paymentCInspect
Finds a payment that already happened for a run, so the work does not run again.
Inputs: find_id (required). Returns: find_id, currency, shortfall, next_action, +18 more
replay_find_decisionsReplay a run's checksARead-onlyIdempotentInspect
Replays the saved condition checks of a run from their recorded evidence, and returns whether the replay gives the same results, with the number of checks replayed. No new search and no new charge.
Inputs: find_id (required). Returns: message, verified, available, decision_root, +3 more
report_wrong_resultReport a result in a listAInspect
Reports a wrong charged result in a list, within 7 days, for review and a refund.
Inputs: check (required), truth (required), record_id (required), request_id (required), +2 more. Returns: status, claim_id, automatic_refund, support_ticket_id
request_account_deletionClose the accountADestructiveInspect
Closes the account. Access ends at once, and the data is erased seven days later.
Inputs: no named parameters. Returns: account_id, grace_days, recoverable, access_revoked, +1 more
research_pricePrice a research reportARead-onlyIdempotentInspect
Fits when the user asks for a written report on a question, before it is approved. Returns the price of one report at each depth (quick, standard and deep) in dollars and in cents, and how many parts of the question each depth researches. A report is charged once, when it is delivered; a run that fails, is cancelled or reads no page is not charged. Nothing runs and nothing is charged.
Inputs: depth. Returns: The price of one report at each depth, charged once on delivery.
resolve_noteResolve a noteBInspect
Marks a note done, or reopens it.
Inputs: note_id (required), resolved (required), expected_version (required), record_id, +2 more. Returns: notes, version
revise_specSave a new versionAInspect
Saves a change to a saved request as a new version. The save fails if someone changed the request after the caller read it.
Inputs: ast (required), spec_id (required), expected_revision_id (required), q, +3 more. Returns: spec_id, archived, revision, account_id, +4 more
revoke_project_memberRemove a project memberBDestructiveInspect
Ends a person's access to a project. Their work stays.
Inputs: account_id (required), project_id (required). Returns: revoked
revoke_spec_linkTurn off a request linkADestructiveInspect
Turns off a shared request link. The request and its results stay.
Inputs: link_id (required), spec_id (required). Returns: link_id, revoked
save_connection_destinationSave a field mappingBInspect
Saves which fact goes into which field of a destination, for every later send.
Inputs: selection (required), connection_id (required). Returns: result
save_relationshipsConnect records by a columnCInspect
Saves the column that links records, such as a city, an owner or a supplier.
Inputs: label (required), field_id (required), collection_id (required), expected_version (required). Returns: label, groups, missing, version, +2 more
save_viewSave a viewBInspect
Saves the filters, columns, sort order and map area of a collection as a personal view.
Inputs: view (required), collection_id (required), expected_version (required). Returns: views, version, collection_id
save_web_source_listSave a source preview as a listAInspect
Saves a preview the user reviewed as an imported list, by its preview hash. The user sees the preview and its explanation before the save. Saving does not research or spend.
Inputs: name (required), import_id (required), preview_id (required), preview_hash (required), +1 more. Returns: id, name, rows, imported, +8 more
save_your_fieldsAdd, rename or remove your fieldsADestructiveInspect
Adds, renames or removes the team's own fields on a collection. Removing a field deletes its values.
Inputs: fields (required), collection_id (required), expected_version (required). Returns: fields, values, version, collection_id
search_webSearch the webAInspect
Fits when the user asks for one fact or a quick answer from the live web, or wants the first rows of a list before the full table is priced. Not for the text of known pages. Fast search: answers in seconds. Returns kind (fact or list), the answer with the page it was read from, rows with each result's name and the facts the request asks for, each value with its page, the search results read (sources), took_ms and request_id. num (1 to 100) sets how many sources come back and page (1 to 10) continues the same list; next_page names the next page. For a list, handoff names the estimated rows and the next steps, a compiled request and a quote for the full table. A call of up to 10 results takes $0.005 from the account's balance; each started block of 10 more results adds $0.006. With speed=fast a call takes $0.001 and a block $0.0012. It needs no quote or approval. A full list is run from a compiled request and its quote, not from this search.
Inputs: q (required), num, news, page, +10 more. Returns: id, kind, news, page, +18 more
search_workspaceSearch your workARead-onlyIdempotentInspect
Fits when the user asks for something already in their SpecStream account, by name or text. Not for a search of the live web. Finds records, lists, requests, projects, people, notes and monitors by name or text.
Inputs: q (required), limit, cursor. Returns: results, next_cursor
send_to_destinationSend records to a toolBDestructiveInspect
Writes the selected results to Google Sheets, Excel, Airtable, Notion, HubSpot, Salesforce or Outlook drafts with the saved field mapping. A CRM updates the records it already has.
Inputs: find_id (required), entity_ids (required), connection_id (required), destination_sha256 (required). Returns: result
set_schedule_deliveryChoose where a monitor reportsADestructiveInspect
Sets where new and changed records go when a monitor finds them: notifications, Slack, a connected tool or a signed webhook. SpecStream keeps this choice on the monitor in the workspace database. Nothing runs.
Inputs: schedule_id (required), expected_version (required), slack, notify, +2 more. Returns: version, delivery, schedule_id, webhook_secret
set_your_fieldSet a team field on recordsAInspect
Sets or clears the value of one team field, such as a status of Called, on the chosen records of a list, kept in the workspace database. Returns the fields and values with their new version. Nothing is charged.
Inputs: key (required), record_ids (required), collection_id (required), expected_version (required), +1 more. Returns: fields, values, version, collection_id
spec_optionsSee what you can ask forCRead-onlyIdempotentInspect
Lists the conditions and facts SpecStream can check for one kind of record.
Inputs: target (required). Returns: bring_back, must_match, entity_type, target_label, +2 more
star_itemStar somethingAInspect
Stars or unstars a request, collection or project, so it shows in the sidebar.
Inputs: item (required), starred (required), expected_version (required). Returns: starred, version
start_researchWrite a research reportADestructiveInspect
Fits when the user wants a written report that answers a question from the live web, with the page behind each cited sentence, and has approved its price. Not for one fact or a quick answer, which the fast web search gives, or for a table of results. Starts a run that splits the question into parts, searches the live web, reads pages and writes a markdown report with numbered citations, or JSON that matches a given JSON Schema. A quick report takes about a minute, a deep one up to about fifteen. Returns the run's id, its status and its price at once. The report is charged once, at delivery, at its depth's price ($0.10 quick, $0.30 standard, $1.00 deep) from the account's balance; a failed or cancelled run is not charged. A depth priced above confirm_max_cents is refused, and the same idempotency_key returns the same run.
Inputs: question (required), idempotency_key (required), confirm_max_cents (required), depth, +5 more. Returns: id, depth, links, status, +2 more
teach_web_sourceTeach a website sourceAInspect
Fits when the user names one public page that holds records and can copy 3 to 8 example rows from it. Not for the text of a page or for a list from the whole web. Reads the page and learns its records from the example rows. Returns a preview and its explanation: how each field is read, which rows are the user's examples and which were inferred, and what happens when the page changes.
Inputs: url (required), name (required), examples (required), exclude, +2 more. Returns: name, rows, count, reuse, +12 more
update_capacity_schedulePause, resume or delete a monitorBDestructiveInspect
Pauses, resumes or deletes a monitor. Missed runs are never bought later.
Inputs: state (required), schedule_id (required), expected_version (required). Returns: object
update_cloud_importStart or stop daily checksBDestructiveInspect
Turns the daily check for new and changed rows on or off for a linked sheet.
Inputs: import_id (required), keepUpdated (required). Returns: result
update_destination_syncStart or stop sendingCDestructiveInspect
Starts or stops continuous sending to a destination.
Inputs: sync_id (required), keepSending (required). Returns: result
update_projectUpdate a projectBInspect
Changes a project's details. The change fails if someone changed the project after the caller read it.
Inputs: project_id (required), expected_version (required), name, monthly_spend_cap_cents. Returns: name, role, version, can_spend, +4 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Live web search, scraping and crawling API for AI. Get structured, LLM-ready data from any site.
Web scraping, search, crawling, structured extraction and research for AI agents.
Crawl, scrape, search the web, and automate browsers at scale with anti-bot bypass.
Web search, browser automation, scraping, crawling and CAPTCHA solving for AI agents.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceProvides live web and document intelligence, including search, company news, scraping, crawling, structured extraction, document parsing, brand intelligence, screenshots, website monitoring, and batch jobs.MIT

OpenSERPofficial
AlicenseAqualityBmaintenanceLive search on six engines (Google, Bing, Yandex, Baidu, DuckDuckGo, Ecosia), image search and single/batch page extraction for AI agents. Keeps source engine and rank. Runs against a self-hosted OpenSERP server or OpenSERP Cloud.3910240 npm3MIT- FlicenseNot gradedqualityAmaintenanceEnables language models and AI assistants to scrape, extract, search, crawl, generate schemas, and monitor web data through an API-backed toolset.-
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to search the live web, fetch clean page content, extract schema-driven structured JSON, check link liveness, detect page changes, run recurring monitors, and verify claims against sources. It offers a free per-IP tier with no API key, plus pay-per-call or prepaid access for higher volume and LLM-backed tools.MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.