What Is MCP (and Why Connect InvesTeam)

The InvesTeam MCP server lets an AI agent drive the same investment pipeline the product drives — brief, clarification, committee, analysis — as a set of callable tools. It speaks the Model Context Protocol, so any MCP-capable client (Claude first) can discover and call InvesTeam's tools directly. The output is decision-support analysis, not investment advice.

What is the Model Context Protocol?¶

The Model Context Protocol (MCP) is an open standard for connecting AI clients to external tools and data. An MCP server publishes a set of named tools; an MCP client — an AI agent, or an application wrapping one — discovers those tools and calls them as part of its reasoning. It is the wiring that lets an agent do something in the world, not just describe it.

InvesTeam ships an MCP server so an agent can run a real investment analysis on a user's behalf: submit a question, answer the analyst's clarifications, and read the committee's findings — all as tool calls, with no glue code on your side.

What is the InvesTeam MCP server?¶

The InvesTeam MCP server is a standalone streamable-HTTP server that sits beside the public API as a second door. It is a thin client of the InvesTeam API: it holds no credentials of its own, no data store, and no internal keys. Each tool call validates its input, forwards your own hfk_ key verbatim to exactly one public /api/v1/** route, and returns that route's JSON unchanged.

Because the server is a projection of the public API, everything true of the API is true here: the same submit→poll pipeline, the same error contract, the same legal disclosure. It is exactly as privileged as the key you present — nothing more. What it adds is ergonomics: an agent gets InvesTeam as tools rather than as HTTP calls it has to construct.

What can an AI agent do with InvesTeam?¶

An agent connected to the InvesTeam MCP server can run the full pipeline as tools. In the shipped tool set it can:

• Create a Master Investment Brief from a free-form question (create_brief).
• Answer the analyst's clarifying questions to sharpen the mandate (answer_clarifications).
• Poll a brief to completion and re-hydrate a past session (get_brief).
• List the user's past sessions (list_sessions).
• Read the committee plan and the final analysis once a committee has run (get_orchestration, get_analysis).
• Confirm which principal and tier a key resolves to before spending a kickoff (auth_whoami).

The live committee-run tools (get_execution, get_transcript) and the convene tool (convene_committee) are gated — they register only when their server-side flags are on, so an agent should not assume they are present. See the MCP Tool Catalog for the exact set and the Agent Patterns guide for the loop.

Should I use MCP or the REST API?¶

Use the MCP server when you are giving an AI agent InvesTeam's analysis as tools — for example, a Claude workflow where a user asks "create an investment brief on this thesis" and the agent runs it. Use the REST API when you are integrating InvesTeam from your own backend code. The two surfaces run the same pipeline against the same hfk_ key; the choice is about who is calling. The API or MCP? page walks the decision.

What InvesTeam over MCP does not do¶

InvesTeam produces decision-support analysis. It is not investment advice, it makes no recommendation tailored to any person, and it has no execution or brokerage surface — there is no tool to place a trade. Every tool that returns AI-generated output carries a legal disclosure your agent must pass through to the end user unchanged.

Next: connect the server to Claude, or read what MCP gives you across any client.