> ## Documentation Index
> Fetch the complete documentation index at: https://ploti.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Your own agent

> An AI agent you build and run yourself, such as a bot on your own server, reads Ploti's records through the CLI, the Python library or the REST API with an API key

An AI agent you build and run yourself might be a bot on your server that watches for new permits or answers your team's questions in chat. It reads the official parcel, zoning, permit and ordinance records Ploti holds for your counties the same way any program does: through the CLI (Ploti's command-line tool), the Python library or the REST API, with an API key. [Keys and setup](https://ploti.ai/docs/developers/keys-and-setup.md) sets each one up.

Using a ready-made assistant such as Muse instead? See [Always-on assistants](https://ploti.ai/docs/mcp/always-on-assistants.md).

**1. Make a key**

A workspace admin makes a key under **Settings**, then **API keys** in
Ploti. Store it as a secret on the agent's machine, as `PLOTI_API_KEY`.

**2. Install the CLI and Ploti's skills**

```bash theme={null}
npm install -g @ploti/cli
ploti skills install <folder your agent reads skills from>
```

The skills are short instruction files that teach the agent Ploti's commands.

**3. Let it work**

The agent runs commands such as `ploti datasets search`, `ploti datasets rows`
and `ploti datasets download`, or calls the same operations from Python.
[Read the data](https://ploti.ai/docs/developers/find-datasets.md) explains each, and [Show results on a map](https://ploti.ai/docs/mcp/maps.md) covers maps.

**4. Hear about changes**

Register a [webhook](https://ploti.ai/docs/mcp/watch-datasets.md) so the agent hears when a
request finishes or a watched dataset changes, so it doesn't have to keep
checking.

## Other ways in

An agent that can fetch a URL and keep a secret can also set itself up from `https://api.ploti.ai/v1/setup.md`, the same instructions the ready-made assistants follow. An agent that speaks MCP can connect to `https://api.ploti.ai/api/mcp` with the key as a bearer token, for results on a Ploti map.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.