FactSet AI-Ready Data

Query FactSet's licensed financial data.

Add to Agentman Studio

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

Resources

What the FactSet AI-Ready Data connector does

The FactSet AI-Ready Data MCP server connects FactSet's licensed institutional financial data — prices, fundamentals, estimates, ownership, filings — to Claude, ChatGPT and any MCP-compatible agent. It exposes 21 read-only tools covering 116 countries and 86,000+ companies. Sign-in is OAuth through FactSet's own login, and what you can actually reach depends on your firm's existing FactSet entitlements.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

FactSet AI-Ready Data tools (21)

  • FactSet_CalendarEvents
  • FactSet_EntityReference
  • FactSet_EstimatesConsensus
  • FactSet_Fundamentals
  • FactSet_FundsETF
  • FactSet_FundsScreener
  • FactSet_GeoRev
  • FactSet_GlobalPrices
  • FactSet_MergersAcquisitions
  • FactSet_Metrics
  • FactSet_Ownership
  • FactSet_People
  • FactSet_PrivateEquityVC
  • FactSet_RBICS
  • FactSet_SupplyChain
  • FactSet_UnstructuredContent
  • FactSet_CompanyScreener
  • FactSet_TermsConditions
  • FactSet_PrivateCompany
  • FactSet_DebtCapitalStructure
  • FactSet_Banks

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • One application scope, and it is all-or-nothing. The server's RFC 9728 descriptor advertised openid, mcp and email on 2026-08-22. Two are identity claims, leaving mcp as the only application scope. No boundary falls between one dataset and another — nothing at consent distinguishes reading prices from reading executive compensation. Entitlements, not OAuth, do that work, and they are configured in FactSet rather than at connect time.
  • What you can reach depends on your contract. Tool availability, packaging and even tool descriptions vary by data tier and firm type. Coverage claims — 116 countries, 86,000+ companies, 16 million+ entities — are FactSet's catalogue maxima, not a guarantee of what your seat returns.
  • The changelog names tools no enumeration confirms. Macroeconomics, InSync Analytics (five tools), Prices and Benchmark Constituents appear in FactSet's release notes but in neither tool list. We could not verify their identifiers.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail, and no readOnlyHint, is published here. Dataset counts come from FactSet's prose documentation, not from JSON Schema.
  • Third-party AI terms may restrict what you build, including a supplier clause against fully autonomous agentic workflows. See the licensing section above. No MCP-specific addendum is published, and one may exist that modifies everything cited here.
  • FactSet documents its own accuracy caveats. Its best-practice guide advises turning off web search "to ensure proper data sourcing", starting a new conversation per topic to avoid token errors, breaking broad questions into targeted ones, and asking for "references, formulas, or explanations of calculations to validate results". It also notes an audit feature on some tools for tracing data back to source. Verify figures before acting on them.
  • Context-window pressure is a documented failure mode. FactSet recommends a smaller model for high-volume queries and sandboxed code execution instead of in-context computation, which indicates large responses are routine rather than exceptional.
  • The RFC 9728 endpoint is an interpolating catch-all. Requests for non-existent resource paths return HTTP 200 with plausible descriptor JSON echoing the requested path back — we confirmed this against two bogus paths, whose bodies differed only in the echoed resource value. No www-authenticate header appeared on the 401, so nothing independently arbitrates which descriptor is canonical. The auth requirement itself is verified by the live 401; treat the descriptor's own path as unconfirmed.
  • A second FactSet MCP server with destructive tools shares the hostname. FPE MCP is a different gateway path, is not this connector, and is not in Anthropic's directory — but FactSet's own FPE setup instructions paste this connector's URL as their example.
  • The vendor's documentation is invisible to most crawlers. The nominated docs URL serves a 3.3 KB JavaScript shell with no server-rendered content; everything cited here came from the portal's underlying content API. Expect search engines and naive fetchers to see nothing there.

Frequently asked questions

How many tools does the FactSet MCP server have?

Twenty-one, and FactSet's own catalogue agrees with Anthropic's listing exactly. But the count understates the surface: each tool takes a dataset selector, and FactSet's documentation enumerates roughly 100 datasets behind those 21 names. FundsETF alone documents 18. Treat 21 as the number of doors, not the number of rooms.

Can the FactSet connector change or delete any of my data?

No. All 21 tools retrieve data and none carries a write, update or delete verb, which matches the Read-only permission Anthropic's directory records. FactSet's own summaries use only retrieval verbs. We could not read the server's tool annotations because the endpoint requires authentication, so this is a names-and-documentation classification.

Why do I see fewer FactSet tools than the 21 listed?

Because your firm's entitlements decide the list. FactSet's version 1.15 changelog states that tool availability and packaging reflect your firm type as well as your data tier, and that Private Company moved to the Premium package. Tool descriptions also adapt to your tier, so two subscribers can see different surfaces from the same server.

Does connecting FactSet to Claude let Anthropic train on FactSet data?

FactSet's supplier terms forbid it for several data providers, and the obligation lands on you. Clauses from DTCC Solutions, RepRisk and ICE Data Services each bar using the data to create, train or fine-tune an AI model. eInforma goes further and prohibits any AI tool that lets its provider reuse user content.

Are fully autonomous FactSet agents allowed?

Not for every dataset. eInforma's rider in FactSet's third-party terms states that fully autonomous agentic workflows operating outside an enterprise governance framework and without human-authorized oversight are not permitted, while workflows configured and governed by a human administrator are. That is a licence term about agent design, not a technical limit the server enforces.

What OAuth permissions does the FactSet MCP server request?

Three, and only one of them means anything. FactSet's resource descriptor advertised openid, mcp and email on 2026-08-22. The openid and email values are identity claims, leaving mcp as a single all-or-nothing application scope. Nothing at the consent screen separates one dataset from another; entitlements do that work instead.

Is the FactSet MCP server the same one that runs Python code?

No, and the distinction matters. FactSet publishes a second server, FPE MCP, on the same mcp.factset.com host under a different gateway path. That one exposes 23 tools including bash command execution and file deletion. This connector's content gateway has neither, and Anthropic's directory lists only this one.

Can the FactSet connector read earnings call transcripts and news?

Yes, through one tool. FactSet_UnstructuredContent searches vectorized CallStreet transcripts, StreetAccount news and EDGAR 10-K, 10-Q and 8-K filings from 2022 onward, returning ranked documents with similarity scores. It is the only tool on the server that reaches unstructured text rather than structured fields.

Sources

More Finance connectors

  • Affinity

    Search your CRM, prep for meetings and update deal records from your AI assistant.

  • Aiera

    Pull earnings-call transcripts, event calendars, SEC filings and company publications into your AI assistant.

  • Airwallex Developer

    Search Airwallex documentation and exercise a sandbox account from an AI agent.

Browse all MCP connectorsClaude and ChatGPT setup guideBrowse AI skills

Product names and logos belong to their owners. Agentman is not affiliated with a connector's maker unless the page says the connector is built and run by Agentman.

Use it in an agent

Put FactSet AI-Ready Data to work.

Connect FactSet AI-Ready Data once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio