Steep MCP Server

Introduction

The Model Context Protocol (MCP) server provides a standardized interface that enables any compatible AI model or agent to access your Steep workspace in a simple, secure way.

Connect to our MCP server natively in Claude, Cursor, Codex, and other clients to query metrics, browse the metric catalog, and explore your workspace, directly from the tools you already use.

What you can do

Through the Steep MCP, an AI agent can read and query your semantic model: the metrics, entities, modules and targets defined in your workspace. An agent can:

• Browse and search the metric catalog, and query metrics for aggregated results.
• List entities and their properties, and query entity rows to work with the individual records behind your metrics.
• Look up modules, target series and target data.
• Search, read and create reports.
• List workspace members and teams.

Because every result comes from the same semantic layer that powers Steep, your AI assistant works with the same trusted, governed metrics as the rest of your organization.

Agents connected through MCP also get the AI context you add to your semantic model, such as synonyms and agent instructions.

Server address

The Steep MCP server is hosted at:

https://mcp.steep.app/mcp

Authentication happens through OAuth. Your agent will only see data you have access to in Steep.

Most clients connect automatically. You add the server URL, sign in, and they handle the OAuth setup for you. A confidential client is an option some platforms can use instead, and a few, such as Gemini Enterprise, require it. See Confidential clients below.

Setup

The MCP server uses streamable HTTP transport and works with any MCP-compatible client. Setup steps differ slightly between clients. The sections below cover the most common ones.

ChatGPT

Setting up the Steep MCP in ChatGPT is a two-step process: a workspace admin adds and publishes the connector once, then anyone in the workspace can enable and use it.

For workspace admins (one-time setup)

1. In ChatGPT, go to Settings → Apps & Connectors → Advanced settings and turn on Developer Mode (if it isn't already).
2. Go back to Apps & Connectors and click Add new connector.
3. Name the connector Steep, enter https://mcp.steep.app/mcp as the MCP Server URL, and set the authentication mode to OAuth.
4. Click Create to authenticate the MCP connection.
5. Go to Apps → Drafts and click Publish next to the Steep app to make it available to your workspace. You will need to review and confirm the security warnings before you can publish.

For everyone else

Once the Steep app has been published in your workspace, enable it in Apps & Connectors and sign in to Steep when prompted.

Claude

Team and Enterprise plans

Setting up the Steep MCP in Claude is a two-step process: an Owner adds the connector once for the organization, then members individually connect and sign in to Steep.

For Owners (one-time setup)

1. In Claude, go to Customize → Connector → Add custom connector.
2. Enter https://mcp.steep.app/mcp as the remote MCP server URL.
3. Click Add to confirm.

For everyone else

Once an Owner has added the Steep connector, go to Settings → Connectors, find Steep in the list, click Connect, and sign in to Steep when prompted.

Pro and Max plans

Individual users can add the connector themselves. Go to Settings → Connectors → Add custom connector, enter https://mcp.steep.app/mcp as the server URL, and sign in to Steep when prompted.

Claude Code

Run the following in your terminal:

claude mcp add --transport http steep https://mcp.steep.app/mcp

Then run /mcp in a Claude Code session to complete the sign-in flow.

Cursor

In Cursor, go to Settings → Cursor Settings → Tools & MCP and add a new MCP server with the URL https://mcp.steep.app/mcp. Or add it directly to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "steep": {
      "url": "https://mcp.steep.app/mcp"
    }
  }
}

After saving, return to Tools & MCP and click Connect to sign in.

Codex

Run in your terminal:

codex mcp add steep --url https://mcp.steep.app/mcp

You'll be prompted to sign in to Steep. If this is your first MCP in Codex, you may need to enable the rmcp feature in ~/.codex/config.toml:

[features]
experimental_use_rmcp_client = true

Other clients

Any MCP-compatible client can connect to Steep using the same server URL. Add the following to your client's MCP configuration:

{
  "mcpServers": {
    "steep": {
      "url": "https://mcp.steep.app/mcp"
    }
  }
}

Refer to your client's documentation for where to place this configuration.

Confidential clients

The clients above connect automatically, registering with Steep the first time they connect. A confidential client is an alternative you register with Steep ahead of time, authenticating with a client ID and secret. Most clients can use one but don't have to. A few platforms support only this method.

Create and manage them in Steep under Settings → Connections → OAuth clients. This requires an admin with permission to manage integrations. To add one, click Create client and set:

1. A name.
2. One or more redirect URIs: the callback URLs the platform you're connecting provides.
3. The scopes the client may use. By default it can use every scope your role allows. Restrict them to limit what the client can access.

Create the client, then copy the client secret. It is shown only once. You'll also get a client ID. Treat the secret like a password. You can revoke a client at any time from the same screen, which immediately invalidates its tokens.

Gemini Enterprise

Gemini Enterprise requires a confidential client. Create one as described above, using this redirect URI: https://vertexaisearch.cloud.google.com/oauth-redirect.

Then add a custom action of type MCP server in Gemini Enterprise, set the authentication mode to OAuth, and fill in the fields:

Then sign in to Steep when prompted.