Agent API

Your own bot can do anything you can in My Board OS. We do not connect QuickBooks, HRIS, or ERP — your agent is the glue. Bring that context in as documents, KPIs, and briefing text, then hold an Executive Review and report back.

Use Help before guessing

If you are unsure how a feature works, search GET /api/v1/help?q= then read GET /api/v1/help/{slug}. Discovery, OpenAPI, MCP, and errors all say the same thing. Never invent endpoints.

Create a token (once)

Sign in on the website and open Account → Agent API.
Name the token (for example OpenClaw) and keep the default scopes unless you need billing or team.
Copy the secret. It is shown once. Send it as Authorization: Bearer … with Accept: application/json.

Coach partners mint a separate token in the coach portal. A personal workspace still needs a subscriber token — there is no “enter workspace” API.

Canonical first run

GET /api/v1 — persona, capabilities, recipes, help.
If unsure, search help.
GET /api/v1/me for org context.
Follow recipe executive-review-from-packet.

Flagship recipe: packet → Executive Review → report back

Gather numbers from tools you already have. Upload exports, optionally patch KPIs, create an Executive Review, poll until pending_approval, read minutes, then approve only if the human asked you to.

Live meetings stall without chair handlers. Prefer Executive Review when acting on the user’s behalf. Queue workers must be running or meetings freeze — same as the boardroom.

OpenClaw / MCP

POST https://myboardos.ai/api/v1/mcp with the same Bearer token. Job-oriented tools wrap the REST API. ChatGPT and Grok can use OpenAPI Actions instead.

{
  "mcpServers": {
    "myboardos": {
      "url": "https://myboardos.ai/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Machine-readable

OpenAPI — also rendered at /developers/openapi
/llms.txt
/.well-known/api-catalog
MCP server card