AI assistants and agents
Use Mapit with AI
Two ways to use Mapit with AI: ask an assistant about your own bike, or have a coding agent write the code that reads it.
Beta, free for now. My Mapit and the service behind it are a beta: they may change or stop, they come with no uptime promise, and Mapit does not charge for them today. Do not rely on them for emergencies or to recover a stolen bike: use the Mapit app. See the beta terms.
Ask an assistant about your bike
My Mapit (Mi Mapit in Spanish) connects an AI assistant to your own Mapit account. Ask “Where is my bike?”, “What rides did I do yesterday?” or “When does my subscription end?”. You sign in with your Mapit account on Mapit's page, and you choose what the assistant may see. It never gets your password.
Plugin or connector
My Mapit comes two ways. The plugin bundles the My Mapit connector with two skills. The marketplace mapit-iot/ai-plugin (a GitHub repository) is added in Claude Code and Codex below. In claude.ai an organization admin adds it under Organization settings > Plugins & skills > Marketplaces, from a private copy (see claude.ai below). If your assistant has no plugins, add the connector by its URL instead, as for claude.ai and ChatGPT below.
claude.ai
In Settings, open Connectors and add a custom connector named My Mapit with the URL https://api.mapit.me/mcp. Then connect it and sign in with your Mapit account.
Team and Enterprise owners can give the plugin to the whole organization: in Organization settings, open Plugins & skills, then Marketplaces, and add the GitHub repository that holds a copy of mapit-iot/ai-plugin. Anthropic accepts only private or internal repositories there, so copy it into a private repository of the organization first. Everyone else adds the connector as above.
ChatGPT
Add a custom connector (an app that uses MCP) named My Mapit with the URL https://api.mapit.me/mcp, and sign in with your Mapit account. Where custom connectors are in ChatGPT depends on your plan and app version.
Claude Code
Install the Mapit plugin from the mapit-iot/ai-plugin marketplace:
/plugin marketplace add mapit-iot/ai-plugin
/plugin install mapit@mapit Then run /mcp, choose plugin:mapit:my-mapit and sign in. The plugin also adds two skills that help the assistant answer questions about your bike and build ride totals.
OpenAI Codex
Install the plugin from the same mapit-iot/ai-plugin marketplace:
codex plugin marketplace add mapit-iot/ai-plugin
codex plugin add mapit@mapit Then sign in with your Mapit account the first time you use it.
Any other MCP client
Add a remote MCP server with the URL https://api.mapit.me/mcp. It uses the Streamable HTTP transport and OAuth sign-in: the client opens Mapit's sign-in page, you approve it, and it keeps a token, never your password.
https://api.mapit.me/mcp What the assistant can and cannot do
- Read-only. It can read your account, vehicles, subscription, where each vehicle last reported, and your routes and ride statistics. It cannot change anything, and it cannot send commands to your bike.
- Location and routes need your tick. On the sign-in page you choose whether to share location and route history. If you leave them unticked, the assistant cannot see them.
- Last reported, not live. A position is what the tracker last sent, and the assistant says when. It cannot follow your bike live.
- Your account only. It sees your vehicles, as the Mapit app shows them.
- You stay in control. See and end every connection at login.mapit.me/connections. It takes effect on the next request.
Claude, ChatGPT and other assistants connected this way are on the Approved apps list. Use is for your own personal, non-commercial purposes, under the API terms of use and the API privacy notice.
Build with an AI coding agent
Give your coding agent these addresses. Each docs page is also available as Markdown: add .md to its address.
https://developers.mapit.me/llms.txt: an index of every page, with a line on each.https://developers.mapit.me/llms-full.txt: all the docs in one Markdown file.https://developers.mapit.me/openapi.json: the OpenAPI description, with every endpoint, parameter and field.
Instructions for the agent
Paste this into your agent's instructions (for example AGENTS.md or CLAUDE.md) or at the start of a task:
You are writing code against the Mapit API (beta).
Docs index: https://developers.mapit.me/llms.txt
All docs in one file: https://developers.mapit.me/llms-full.txt
OpenAPI spec: https://developers.mapit.me/openapi.json
- Base URL: https://api.mapit.me/v1
- Auth: send a personal key as "Authorization: Bearer $MAPIT_KEY". Read the key
from an environment variable. Never hard-code it, print it or log it.
- Rate limit: 30 requests a minute per key or access token, for the REST API and the MCP connector. Bursts above it may be tolerated, never rely on them. On 429, wait Retry-After seconds.
Poll /status at most once a minute. See https://developers.mapit.me/limits.md
- Every 200 answer has an ETag: send it back in If-None-Match. 304 means
nothing changed.
- Version 1 only gains fields: ignore fields you do not know.
- Ids and cursors are opaque: store them, never build or parse them.
Follow nextCursor until it is null.
- Positions are last reported, not live: show when (lastLocationAt).
- Location and routes need permission on the key: 403 means it lacks it.
- Read-only, except disconnecting apps and keys. Never call DELETE unless
the user asks.
- Personal, non-commercial use: the user's own vehicles only.
- Python: use the aiomapit package instead of raw HTTP. Python: aiomapit
For Python, the fastest path is aiomapit, Mapit's official async client (Python 3.12 or newer). It is typed, handles paging and is what the Home Assistant integration uses. Create a personal key first.
pip install aiomapit import asyncio, os
from aiomapit import MapitClient
async def main():
async with MapitClient(os.environ["MAPIT_KEY"]) as client:
async for status in client.iter_statuses():
if status.state:
print(status.vehicle_id, status.state.latitude, status.state.longitude)
asyncio.run(main()) Other languages can use the reference or the OpenAPI file to generate a client.
