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

# MCP overview

> Use Blueprints from MCP-compatible AI clients.

Blueprints provides a remote Model Context Protocol server for AI clients that support HTTP MCP servers.

## Endpoint

```text theme={null}
https://mcp.useblueprints.ai/mcp
```

The MCP server uses your scoped Blueprints API key to act on your account.

```http theme={null}
Authorization: Bearer bp_YOUR_API_KEY
```

## When to use MCP

Use MCP when the operator is working inside an AI client and wants account-aware assistance. The MCP server gives the client structured tools for reading blueprint state, making edits, validating changes, and controlling deployments.

Use the HTTP API instead when you are writing a deterministic integration, service, or backend job.

## What MCP is for

Use MCP when you want an AI client to help with blueprint authoring and account-aware workflows:

* Discover or create blueprints
* Read the current blueprint state
* Add, configure, connect, and arrange blueprint building blocks
* Search supported market inputs
* Deploy or stop a blueprint after review

The MCP server does not bypass product safeguards. Actions still run as your authenticated account and are subject to the same validation and permissions as the API.

Grant only the scopes your MCP client needs. A full blueprint-building client usually needs `blueprints:read`, `blueprints:write`, and `blueprints:deploy`. Omit `blueprints:deploy` for clients that should inspect and edit drafts but never start or stop live automation.

## Operating expectations

MCP tools can mutate account-owned blueprints. Review tool calls in your client before approving edits, deploys, or stops.

Good MCP prompts are specific about the target blueprint and desired outcome. For example, ask the client to inspect the current blueprint, describe the planned edits, validate the result, and wait for approval before deployment.