Documentation

API & MCP

Build on top of gehirn.md. A REST API and an MCP server expose COMPANY.md generation, tone-of-voice analysis, and design-system markdown. Choose a generation type below for the endpoints and tools that apply.

Rate limits

5 generations per IP per 10 minutes. Cache reads for COMPANY.md are always free and never counted.

REST & MCP

Toggle the artifact you need — each mode lists its HTTP routes and the MCP tools that map to it.

REST — COMPANY.md

Async generation with cache and job polling. Same 5 generations per IP per 10 minutes rule; cache reads stay free.

POST /api/v1/generate

Send a URL, get markdown back. A cache hit returns 200 immediately. A fresh generation returns 202 with a jobId to poll. Pass force: true to skip the cache.

curl -X POST 'https://company-md.com/api/v1/generate' \
  -H 'Content-Type: application/json' \
  -d '{"url":"https://example.com"}'

GET /api/v1/jobs/:id

Poll until status is done. The content field will be set.

curl 'https://company-md.com/api/v1/jobs/<jobId>'

GET /api/v1/cache/:domain

Fetch cached markdown for any domain. Append ?format=json for JSON.

curl 'https://company-md.com/api/v1/cache/example.com'

MCP — COMPANY.md

Endpoint https://company-md.com/api/mcp. Tools:

generate_company_md — generate or return cached markdown
get_cached_company_md — hostname lookup
get_job_status — job polling (same as REST)

Cursor

{
  "mcpServers": {
    "company-md": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://company-md.com/api/mcp"]
    }
  }
}

MCP reference

Point any MCP client at the streamable HTTP endpoint (includes all tools below).

https://company-md.com/api/mcp

generate_company_md — COMPANY.md from a homepage URL
get_cached_company_md — cached markdown by hostname
get_job_status — async job status
analyze_tone_of_voice — brand-voice markdown + profile preview
analyze_design_system — DESIGN.md markdown + profile

Claude Desktop

Use the same mcp-remote config as Cursor. See the Anthropic docs for where to add it.