AIpplied Labs for agents

AIpplied Labs is an AI consulting company based in the Chicago area. This page describes the ways an AI agent can talk to it. Nothing here needs an account or an API key.

Every answer these endpoints return comes from the company's own written copy. A small decision model (Jev) picks which written answer fits a question, and it never writes text of its own. When nothing fits, you get a short fallback instead of a guess.

MCP server

To add it to Claude Code, run:

claude mcp add --transport http aipplied-labs https://www.aippliedlabs.com/mcp

Tools:

A2A agent

Send a text message with a question, and the reply is a single message with the answer. The agent keeps no tasks.

JSON endpoints

Both are described in /openapi.json.

POST /api/ask

Request:

{"question": "How long does a build take?"}

Response: matched is true when a written answer fits, and paragraphs holds the answer.

POST /api/check

Request:

{"business": "trades", "team": "2-5", "text": "Chasing quotes on Sundays", "ticks": ["quotes-sit"], "hours": 6}

Response: systems holds up to three systems, each with name, said, get, and live. An empty list means nothing clearly fits.

GET /api/health

Returns {"status": "ok"} when the endpoints are up.

Limits and auth

Anonymous callers get 12 questions and 8 checks per minute per IP address. To get your own limit instead of sharing one per IP, get a free anonymous token. The steps are in /auth.md.

In the browser (WebMCP)

The homepage registers the same tools with navigator.modelContext in browsers that support WebMCP. It also registers request_working_session, which fills in the booking form for a person to review and send.

Discovery