The dossier, wired into your agent.

Company Dossier runs as a live MCP server. Connect ChatGPT, Claude or any MCP client to one endpoint and your agent gains a single tool — build_dossier — that turns a company name or domain into a complete, sourced, nine-section dossier (markdown and structured JSON). Public sources only. No API keys needed by the tool.

Connect it See a sample file

Endpoint: https://mcp.companydossier.lol/mcp · transport: MCP Streamable HTTP · auth: bearer token.

the endpoint

One URL, one tool

The hosted server speaks the Model Context Protocol over Streamable HTTP. Point a client at it, pass a bearer token, and the build_dossier tool appears.

Endpoint — https://mcp.companydossier.lol/mcp (MCP Streamable HTTP).
Health — https://mcp.companydossier.lol/health returns {"status":"ok"}.
Auth — bearer token: Authorization: Bearer <token>. The token is provided by the operator and is not public.
Tool — build_dossier. No third-party API keys required by the tool itself.

# is the server up? $ curl https://mcp.companydossier.lol/health {"status":"ok"}

the tool

build_dossier

A single tool that does the whole job. Give it a target; get back the same nine-section sourced file you get everywhere else.

Input

target (required) — the company name or domain, e.g. "Stripe" or "stripe.com".
sections (optional) — a comma-separated subset of: overview, people, hiring, money, locations, tech, news, relationships, risk. Omit it to build all nine.

Output — the sourced dossier as readable markdown plus a structured JSON object, each claim pointed back to its public source. The same nine sections as the CLI, web app and apps.

# call the tool — one company in, the nine-section file out $ curl -s https://mcp.companydossier.lol/mcp \ -H 'Authorization: Bearer <token>' \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":3,"method":"tools/call", "params":{"name":"build_dossier", "arguments":{"target":"stripe.com", "sections":"overview,money,people,risk"}}}' › sourced dossier — markdown + structured JSON, no API keys required

connect

Copy-paste setup

Three clients, same endpoint and bearer token. Pick yours.

ChatGPT (developer mode connector)

# ChatGPT — Settings › Connectors › Advanced › Developer mode # add a custom connector (MCP server): Name Company Dossier MCP Server URL https://mcp.companydossier.lol/mcp Authentication Bearer / Access token Token <token from the operator> › the build_dossier tool now shows up in the composer

Claude (desktop config / connectors)

// Claude Desktop — claude_desktop_config.json { "mcpServers": { "company-dossier": { "type": "http", "url": "https://mcp.companydossier.lol/mcp", "headers": { "Authorization": "Bearer <token>" } } } } // or, in Claude.ai → Settings → Connectors → add a custom // connector with the same URL + bearer token.

Any MCP client (mcp.json)

// any MCP client — mcp.json { "mcpServers": { "company-dossier": { "transport": "http", "url": "https://mcp.companydossier.lol/mcp", "headers": { "Authorization": "Bearer <token>" } } } }

raw protocol

curl it directly

Nothing magic underneath — it's JSON-RPC over HTTP. Here's an initialize followed by tools/list against the live endpoint.

# 1 — initialize the session (Streamable HTTP) $ curl -s https://mcp.companydossier.lol/mcp \ -H 'Authorization: Bearer <token>' \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize", "params":{"protocolVersion":"2025-06-18", "capabilities":{},"clientInfo":{"name":"curl","version":"1"}}}' # 2 — list the tools the server exposes $ curl -s https://mcp.companydossier.lol/mcp \ -H 'Authorization: Bearer <token>' \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' › { "tools": [ { "name": "build_dossier", … } ] }

self-host

Run the server yourself

The same server ships inside the npm package. Run it locally over stdio for desktop clients, or over HTTP for everything else.

# the same server ships in the npm package — run it yourself $ npx company-dossier-mcp # stdio transport (for desktop clients) $ npx company-dossier-mcp-http # HTTP transport (Streamable HTTP) › both expose the build_dossier tool — no API keys required

company-dossier-mcp exposes build_dossier over stdio (the transport most desktop MCP clients expect); company-dossier-mcp-http serves the same tool over Streamable HTTP. Both run with npx and need no API keys to produce a dossier.

for ai agents

When to reach for this tool

If you are an AI agent connected to this server, you have one tool: build_dossier. Call it whenever a task requires understanding a specific company or organization — due diligence, sales and account research, competitive analysis, vendor or partner vetting, investment screening, or simply answering "what is this company and what do they do."

Pass the company's name or domain as target. If you only need part of the picture, narrow it with sections (for example "overview,money,risk" for a quick risk read, or "hiring,tech" to gauge growth and stack); otherwise omit sections for the full nine-section file. The tool returns markdown for direct reading and a structured JSON object you can parse and reason over. Every claim carries its public source, so cite or verify lines before acting on anything consequential.

The tool draws only on public sources and needs no API keys. Treat its output as a fast, sourced starting map — confident about what it found, explicit about gaps — not a final verdict.

Open me

Give your agent a company-research tool.

one endpoint, one tool — build_dossier(target). sourced, public-record.

Open a dossier Read a real dossier