---
url: https://docs.ridgedata.ai/guide/mcp.md
---
# Use Ridge in Claude and ChatGPT

Our MCP server connects Ridge to AI assistants such as Claude and ChatGPT. Once it is added,
you can find data, build, change, and explore ridges (dashboards), and publish them, all
from a chat.

The MCP URL is:

```text
https://app.ridgedata.ai/api/v1/mcp
```

## Claude

Claude adds Ridge as a custom connector. A connector added on claude.ai is also available in
the Claude desktop and mobile apps.

1. In Claude, open **Customize** → **Connectors** and choose **Add custom connector**. The
   **Add to Claude** button on your Ridge home page opens this form with the details filled in.
2. Enter `Ridge AI` as the name and the MCP URL above as the server URL, then click **Add**.
3. Click **Connect** and sign in to Ridge.

On Claude Team and Enterprise plans, an owner may need to add the connector for the
organization first.

## ChatGPT

ChatGPT adds Ridge as a custom MCP server.

1. Open [Plugins](https://chatgpt.com/plugins) in ChatGPT.

2. Click **Add** and select **Add custom MCP server**.

3. Fill in the fields:

   | Field          | Value                                     |
   | -------------- | ----------------------------------------- |
   | Name           | `Ridge AI`                                |
   | Description    | `Build and explore Ridges from your data` |
   | Server URL     | `https://app.ridgedata.ai/api/v1/mcp`     |
   | Authentication | OAuth                                     |

   The icon is optional. You can download it from the **Add to ChatGPT** dialog on your Ridge
   home page.

4. Save the server and sign in to Ridge.

## Claude Code

Add the MCP from your terminal:

```bash
claude mcp add --transport http ridge-ai https://app.ridgedata.ai/api/v1/mcp
```

Then run `/mcp` in Claude Code, select **ridge-ai**, and choose **Authenticate** to sign in
to Ridge.

Claude Code can also upload local files as datasets; see [Upload data](#upload-data).

## Signing in

Adding the MCP opens Ridge's sign-in. Choose the organization the assistant works in, then
approve what it may do:

* **See ridges, datasources and connections.** Always granted.
* **Build and edit ridges.**
* **Publish ridges and manage their embeds.**
* **Upload datasets and replace their data.**

You can untick any of the last three. The assistant then doesn't get the tools that need them.
The MCP never deletes ridges or data.

Assistants mark each tool as read-only or as one that changes Ridge, so you can approve them
in groups. Replacing an uploaded dataset's data is marked destructive, so the assistant asks
before it does it.

## What you can ask for

* **Find data and ridges.** List your organization's datasets and ridges.
* **Show a ridge.** The ridge appears in the chat with live charts. Expand it to fullscreen,
  or click **Open in Ridge** to open it in Ridge.
* **Build a ridge.** Describe what you want to see. The build runs on Ridge, the assistant
  answers the build's questions and asks you only for what it can't infer. The result is saved
  as an unpublished draft, which the assistant shows you.
* **Change a ridge.** Ask for a change to an existing ridge. It runs as the ridge's next
  version and is shown the same way.
* **Publish a ridge.** Publishing the draft makes it visible to your organization's viewers.
* **Embed a ridge.** Ask for an embed of a published ridge to get a signed URL and the
  `<iframe>` HTML to add it to your site. See [Iframe embed](./iframe).

Builds and changes keep running if you close or reload the chat. Ask the assistant to check
on them when you return.

## Upload data

The assistant can upload a file as a new dataset, or replace an uploaded dataset's data. It
needs a shell that can reach Ridge, as Claude Code has. Chat apps whose sandbox blocks
outside hosts can't upload, so add the data in Ridge instead.

Uploads are single Parquet files of up to 64 MiB. The assistant converts CSV or JSON to
Parquet first.

## Disconnect

Connected assistants are listed under **Connected Apps** on your
[Ridge account page](https://app.ridgedata.ai/account). Disconnecting one signs it out at
once. You can also remove the connector in the assistant itself.

## Limits

* Ridges whose data is transformed by a semantic model can be shown but not changed from a
  chat.
* Embeds created from a chat are iframe embeds. Set up a [JavaScript embed](./javascript) in
  Ridge.
