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

# Research Desk MCP Server

> Connect an AI assistant to Research Desk to find, create, and configure projects and surveys using natural language.

The Research Desk MCP Server lets you work with your Research Desk account through an AI assistant. Instead of clicking through the platform UI or building an API integration, you can ask an assistant in Claude, ChatGPT, or any other MCP-compatible AI tool to look up projects, check survey status and fielding performance, and set up or update research projects in Research Desk.

It is built on the [Model Context Protocol (MCP)](https://modelcontextprotocol.io), an open standard for connecting AI assistants to software platforms.

<Note>
  The MCP server is included with your Research Desk account at no additional charge. Contact your account manager or [clientsuccess@repdata.com](mailto:clientsuccess@repdata.com) to get started.
</Note>

## Requirements

* A Research Desk account. Any Research Desk user can connect. There are no separate credentials, API keys, or provisioning steps needed.
* An MCP-compatible AI assistant. Verified with **Claude** and **ChatGPT**; works with any client that supports remote MCP servers with OAuth.

## Connect your assistant

Add this server URL to your assistant as a custom connector:

```
https://mcp.researchdesk.com/mcp
```

<Steps>
  <Step title="Add the connector">
    In your assistant, add a new MCP server / custom connector and paste the URL above. In company workspaces this step is typically done once by an admin:

    * **Claude** — a Team/Enterprise Owner adds it under Organization settings → Connectors. See [Anthropic's custom connector guide](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp).
    * **ChatGPT** — a workspace admin enables developer mode and adds it as an MCP app. See [OpenAI's developer mode guide](https://help.openai.com/en/articles/12584461-developer-mode-and-mcp-apps-in-chatgpt).
  </Step>

  <Step title="Sign in">
    Your assistant opens a browser window. Sign in with your normal Research Desk credentials and approve the connection. Each person connects using their individual credentials.
  </Step>

  <Step title="Verify">
    Ask your assistant to list your Research Desk projects. If it returns your projects, you're connected.
  </Step>
</Steps>

## How authentication works

Sign-in uses standard OAuth 2.0 — your assistant handles the whole flow and refreshes tokens automatically. Three things follow from this:

* **Your Research Desk permissions apply.** The assistant can see and change only what your Research Desk role allows in the UI, and all activity is attributable to you.
* **The endpoint is shared.** There are no per-customer URLs; access is determined entirely by who signs in.
* **Tokens lapse.** If a working connection starts failing with authentication errors, disconnect and reconnect the connector to force a fresh sign-in.

## Developers and custom clients

The server speaks MCP over streamable HTTP. From Claude Code:

```bash theme={null}
claude mcp add --transport http researchdesk https://mcp.researchdesk.com/mcp
```

For other clients, point them at the URL with no pre-configured credentials — the server advertises its authorization server via OAuth protected-resource metadata (`https://mcp.researchdesk.com/.well-known/oauth-protected-resource/mcp`) and clients run authorization code + PKCE from there. Request the `offline_access` scope for refresh tokens.

<Warning>
  The server supports client-ID metadata documents but not Dynamic Client Registration. Clients that require DCR cannot complete sign-in.
</Warning>

For systematic, high-volume, or scheduled integrations, use the [Research Desk Demand API](/demand/overview/00-getting-started) instead. The MCP server is designed for assistant-driven, human-in-the-loop work.

## Next steps

<CardGroup cols={2}>
  <Card title="Tools" icon="wrench" href="/mcp/tools">
    What the server can do, and what to know before writing changes.
  </Card>

  <Card title="FAQ" icon="circle-question" href="/mcp/faq">
    Pricing, safety, limitations, and troubleshooting.
  </Card>
</CardGroup>