> For the complete documentation index, see [llms.txt](https://docs.chirpwireless.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.chirpwireless.io/api/mcp-server.md).
# MCP Server
Connect your favorite AI app straight to your Chirp home with MCP — sign in, no API key needed.
You can ask the AI app you already use to help set up your Chirp home. Connect it through **MCP**, then ask it to add a sensor, create an alert, or build an automation. It can make those changes in your home rather than just describe the steps.
**MCP (Model Context Protocol)** is the connection that lets an AI client find and call Chirp's operations. Sign in with your Chirp account and work within the home and permissions you already have. Reading sensor history is available alongside the setup work.
MCP (Model Context Protocol) is simply an agreed way for an AI app to reach out and use something else safely. Chirp speaks it, so the app you already use can connect to your home without any custom plumbing on your side — [Claude Code](https://claude.com/claude-code), Claude Desktop, ChatGPT, Codex, Cursor, and others all support it.
Because MCP is a shared standard rather than a deal with one company, that list keeps growing. If your AI app supports MCP, it can talk to your home.
The best part: there's no key to generate, copy, or paste. You sign in with your usual Chirp account in your browser, and the connection carries exactly your own permissions — nothing more. If you can't see something in the Chirp app, your AI app can't see it either.
## What address do I connect to?
```
https://mcp-auth.chirpwireless.io/mcp
```
The connection uses Streamable HTTP, which the clients below handle for you — you just paste the address.
## How do I connect Claude Code to Chirp?
1. Add Chirp as a connection:
```
claude mcp add --transport http chirp https://mcp-auth.chirpwireless.io/mcp
```
2. Run `/mcp` inside Claude Code. Your browser opens for sign-in.
3. Sign in with your usual Chirp account and approve the connection.
4. Back in Claude Code, `/mcp` now shows Chirp as connected — and the tools are there. Try asking it which of your sensors went quiet this week.
## How do I connect Claude Desktop to Chirp?
Add Chirp as a custom connector in Claude Desktop, using `https://mcp-auth.chirpwireless.io/mcp` as the server address, then sign in with your usual Chirp account when your browser opens. Menu names change between releases, so follow [Anthropic's guide to custom connectors](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp) for the current steps.
## How do I connect Grok to Chirp?
Use Grok to ask about your Chirp home and switch a connected plug or light on or off. Connect it from your phone or computer using your usual Chirp sign-in.
1. Open [Grok Connectors](https://grok.com/connectors) and sign in to your Grok account.
2. Choose **New Connector**, then **Custom**.
3. Paste `https://mcp-auth.chirpwireless.io/mcp` into the server URL field.
4. Follow the sign-in prompt, use your usual Chirp account, and approve the connection. There is no Chirp or xAI API key to enter.
5. In a Grok conversation, ask *“Use Chirp to tell me which home is selected and list my devices.”* Make sure it sees the home you intended.
6. Ask about a reading, such as *“What is the living room temperature, and how recent is it?”* To try control, ask which commands are saved for a named plug or lamp, then ask Grok to run one.
[Grok's connector guide](https://docs.x.ai/grok/connectors) describes the custom connection flow. If you use a Grok Business or Enterprise account, your team administrator must add the connector first; see [the administrator instructions](https://docs.x.ai/grok/connector-management).
To use the connection by voice in the car, see [Use Chirp from Your Tesla](/ai-assistant/tesla-grok.md).
## How do I connect a different AI app to Chirp?
Other apps that support remote MCP servers, such as ChatGPT, Codex and Cursor, can use the same address. Each one puts the setting in a different place and changes it from time to time, so follow that app's own guide to adding a remote MCP server. If it asks for a transport, choose **Streamable HTTP**.
Whichever app you use, the rest is identical: your browser opens, you sign in as yourself, and it sees exactly what you see. Your app's own documentation will tell you where its MCP setting lives.
## How do I choose which home my AI app sees?
If you're only in one home, skip this — everything just works.
The plain `/mcp` address follows whichever home is currently selected in the Chirp app. Switch homes in the app and reconnect, and your AI client sees the new one.
If you'd rather pin the connection to a specific home — handy if you look after a holiday flat as well as your own place, and want each AI client fixed to one — use the address with the home's ID in it:
```
https://mcp-auth.chirpwireless.io/o/{organizationId}/mcp
```
You'll find the ID for a home in the Chirp web app. You can only connect to a home you're a member of; anything else is refused.
## What can my AI app do with Chirp?
Once connected, your AI client can work across your home on your behalf. Here's what it can reach:
| Area | What your AI app can do |
| ----------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Sensors** | List the sensors in your home, see their details, and add a new one for you — a LoRaWAN sensor or a tracker — picking from the available sensor profiles. |
| **Switching things on** | See what a device is set up to do, do it, and check the device actually got the message — so *"turn on the lamp"* works from the app you're already in. |
| **Pretend sensors** | Set up an emulated sensor, change what it reports, send a one-off reading to test an alert, and move a sensor between pretend and real. It can put any of your real sensors onto the emulator, and take a pretend one live onto a LoRaWAN connection. |
| **Connections** | See how your home is connected and set up a new connection for a sensor. |
| **Automations** | Go through the automations you've built and tell you what each one does. |
| **Alerts** | Look at your alerts, summarize what's been triggering and how often, clear one that you've dealt with, and set up which alerts reach you and where. |
| **Dashboards** | List your dashboards and pull the data behind a widget, so it can chart or explain a reading you're looking at. |
| **History** | Look back over when your alerts fired and when your automations ran, and map out which sensor is measuring what. |
| **Household** | Look up your home's details and members, invite someone new, and set what they're allowed to see. |
Everything it does runs with your permissions. For destructive or important actions, Chirp supplies safety information that a compatible AI app can use to ask before it continues.
## How does my AI app know what an action does?
Chirp publishes a readable name, description, and safety notes with each action. AI apps that support these notes can use them to decide whether to run an action immediately or ask you to confirm.
| Safety note | Examples | What it tells the app |
| -------------------------------- | ------------------------------------------------------------------------------ | ---------------------------------------------------------------------------------- |
| **Only reads** | List sensors, read alert history, get the data behind a widget | The action does not change your home. |
| **Changes or removes something** | Remove a sensor, switch a device, change a connection, update alert recipients | The action can affect your home or equipment, so the app can ask for confirmation. |
| **Looks outside your home** | Search a product catalog or the web for hardware | The action accesses information beyond your Chirp data. |
Creating a sensor or dashboard changes your home, but it does not overwrite or switch off anything already there and can be reversed by deleting the new item. Chirp therefore does not describe creation as destructive.
These notes help the AI app explain an action; they are not permission checks. Each app decides how to present confirmations. Chirp still enforces your account permissions, so the app cannot do anything that your account is not allowed to do.
## How is the MCP server different from the AI assistant in Chirp?
Chirp’s built-in [AI Assistant](/ai-assistant.md) works like an experienced integrator by your side. It can add sensors, build automations, and configure alerts from the home project you describe. Open it using **AI Chat** in the sidebar; you can start without connecting an external AI client.
MCP is for when you'd rather bring your own AI app to the same home — because you live in your desktop client all day, because you want your home's data in the same conversation as everything else you're working on, or simply because you prefer that assistant. Both talk to the same home and respect the same permissions.
## How is the MCP server different from the API?
The [REST API](/api/public-rest-api.md) is for code *you* write — a script that exports readings into a spreadsheet every Sunday, or a bridge to another home platform. It runs on a key you create in [Settings → API Keys](/settings/api-keys.md), with exactly the scopes you give it.
MCP is for an AI client acting **as you**, in a conversation, with your own sign-in and your own permissions. No key to manage, and nothing to write. If you want something to run unattended on a schedule, use the API. If you want to ask questions and get things done in the moment, use MCP.