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

# Evermuse MCP Server

> Connect Claude, ChatGPT, Cursor, Codex and other AI agents to your Evermuse customer evidence and product knowledge over the Model Context Protocol.

The Evermuse MCP server gives your AI agent the customer understanding behind better product decisions. It connects any [Model Context Protocol](https://modelcontextprotocol.io) client to your Evermuse workspace, so the agent can research customer needs, pain points and feedback, read the meetings and documents behind them, and draft PRDs, specs, briefs, personas and user stories grounded in what customers actually said.

Evermuse turns scattered customer signals into product understanding that grows with your business. The MCP server brings that understanding into the tools where you already work.

## Server URL

```text theme={null}
https://api.evermuse.com/api/mcp
```

The server is a hosted, remote MCP server. It uses the Streamable HTTP transport and authenticates with OAuth 2.1, so you sign in with your Evermuse account. You don't need an API key or a local install.

## What your agent can do

<CardGroup cols={2}>
  <Card title="Research customers" icon="magnifying-glass">
    Search customer needs, feedback, quotes and pain points across every meeting, call, email and document in your
    workspace, with links back to the original evidence.
  </Card>

  <Card title="Find the important patterns" icon="chart-network">
    Connect findings across sources to uncover recurring pain points, machine-suggested opportunities and competitor
    capabilities.
  </Card>

  <Card title="Turn evidence into action" icon="file-lines">
    Run guided workflows that draft PRDs, feature specs, briefs, personas and user stories grounded in cited customer
    evidence.
  </Card>

  <Card title="Pick up where you left off" icon="floppy-disk">
    Save meetings, notes, documents and shaping notes back to Evermuse, so later conversations start from what you already know.
  </Card>
</CardGroup>

## How it works

<Steps>
  <Step title="Connect">
    Add `https://api.evermuse.com/api/mcp` as a custom connector or MCP server in your AI client. See
    [Connect your AI client](/mcp/connect).
  </Step>

  <Step title="Sign in and choose a workspace">
    Your client opens the Evermuse sign-in page. Sign in or create an account, pick the workspace the agent should
    access, and approve the connection. See [Authentication](/mcp/authentication).
  </Step>

  <Step title="Ask a product question">
    Ask things like *"What are the biggest pain points in our customer interviews?"* The agent loads Evermuse's research
    methodology, picks your product, runs the matching workflow and answers with citations that link to the original
    evidence in Evermuse. See [Example prompts](/mcp/examples).
  </Step>
</Steps>

## Supported clients

Evermuse works with any MCP client that supports remote servers over Streamable HTTP with OAuth. Step-by-step setup is
available for:

| Client | Connection method |
| - | - |
| Claude (claude.ai, Desktop, mobile) | Custom connector |
| Claude Code | `claude mcp add --transport http` |
| ChatGPT | Custom connector (Apps & Connectors) |
| Codex | `~/.codex/config.toml` + `codex mcp login` |
| Cursor | MCP settings or `mcp.json` |
| VS Code (GitHub Copilot agent mode) | `mcp.json` |
| Any other OAuth 2.1 MCP client | Point it at the server URL |

## At a glance

| Property | Value |
| - | - |
| Server name | `Evermuse` |
| Endpoint | `https://api.evermuse.com/api/mcp` |
| Transport | Streamable HTTP, stateless, JSON responses (no SSE stream) |
| Authentication | OAuth 2.1 with PKCE (S256), Dynamic Client Registration, Client ID Metadata Documents; API keys also supported |
| Capabilities | Tools and prompts (skills). Resources are used only by early-access MCP Apps tools. |
| Tools | 27 tools: 22 read-only and 5 that write to your workspace. See [Tools reference](/mcp/tools). |
| Data access | Only the single workspace you authorize. OAuth connections expose only Evermuse tools. |
| Rate limit | 1,000 requests per hour per access token |
| Pricing | Included with Evermuse. Each successful tool call uses one Evermuse Credit (`find_skills` and `read_skills` are free). See [Credits](/mcp/limits-and-errors#credits). |

## Requirements

* An Evermuse account. You can create one during the connection flow with Google, Microsoft or email.
* An Evermuse workspace with at least one product. The sign-in flow walks you through creating both if you don't have
  them yet.
* An MCP client that supports remote (HTTP) servers with OAuth.

<Note>
  All research data in Evermuse belongs to a **product** inside your workspace. Agents should start by calling
  `get_products` and passing the chosen `product_id` to every product-scoped tool. The server's built-in instructions
  tell agents to do this automatically.
</Note>

## Next steps

<CardGroup cols={2}>
  <Card title="Connect your AI client" icon="link" href="/mcp/connect">
    Setup instructions for Claude, ChatGPT, Cursor, Codex, VS Code and others.
  </Card>

  <Card title="Tools reference" icon="screwdriver-wrench" href="/mcp/tools">
    Every tool, its parameters, annotations and what it returns.
  </Card>

  <Card title="Example prompts" icon="message" href="/mcp/examples">
    Real-world prompts and the tool calls they trigger.
  </Card>

  <Card title="Security & privacy" icon="shield-check" href="/mcp/security-and-privacy">
    How access is scoped, what is stored and what is logged.
  </Card>
</CardGroup>


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