Dovetail MCP Server

Official Dovetail MCP server. Gives AI agents read access to projects, insights, highlights, transcripts, and search across your Dovetail customer research workspace.

Collaboration by Dovetail OAuth2 active

Connect (Remote Server) GitHub Docs

Overview

The Dovetail MCP server connects AI assistants like Claude, ChatGPT, and Cursor to your Dovetail customer intelligence workspace. It acts as a secure read-only bridge that lets agents query research projects, pull insights and highlights, fetch interview transcripts and data points in markdown, and search across the entire workspace using natural language. Permissions follow the authenticated user, so the AI only sees projects and data the user already has access to and cannot make any changes to Dovetail.

The server is officially built and maintained by Dovetail and is offered in two flavors. A hosted Streamable HTTP endpoint at https://dovetail.com/api/mcp supports OAuth 2.1 and bearer token auth and works with clients like Cursor, Windsurf, ChatGPT, Figma Make, and Microsoft Copilot Studio. A self-hosted STDIO version (the dovetail/dovetail-mcp Node.js project on GitHub) is available for clients that only support local servers, such as Claude Desktop. Both expose the same set of read tools for projects, insights, highlights, data, and workspace search.

Tools

Tool	Description
`get_dovetail_projects`	List projects available in the connected Dovetail workspace.
`get_project_highlights`	Retrieve highlights tagged within a specific Dovetail project.
`list_project_insights`	List insights belonging to a given project.
`get_project_insight`	Fetch metadata for a specific insight by ID.
`get_insight_content`	Return the full insight body rendered as markdown.
`list_project_data`	List data entries (notes, transcripts, files) within a project.
`get_project_data`	Retrieve metadata for a specific data entry by ID.
`get_data_content`	Return the data entry (e.g. interview transcript) as markdown.
`search_workspace`	Search across all projects, insights, highlights, and data in the workspace.

Setup Guide

Option A: Hosted server (recommended)

Most MCP clients (Cursor, Windsurf, ChatGPT, Figma Make, Microsoft Copilot Studio) can connect directly to the hosted Streamable HTTP endpoint with OAuth 2.1.

Server URL:

https://dovetail.com/api/mcp

In Cursor or another HTTP-capable client, add the server URL and complete the browser-based OAuth login when prompted. ChatGPT and Figma Make also have first-party Dovetail integrations.

Option B: Self-hosted (STDIO) for Claude Desktop

For clients that only support STDIO (such as Claude Desktop), run the official Node.js server locally.

Prerequisites:

Node.js 22 or higher
A Dovetail Personal API token from your account settings (see the Dovetail API docs)

Build from source:

git clone https://github.com/dovetail/dovetail-mcp.git
cd dovetail-mcp
yarn install
yarn build

Or download the prebuilt index.js from the latest GitHub release.

Add to your MCP client config (e.g. Claude Desktop claude_desktop_config.json):

{
  "mcpServers": {
    "dovetail-mcp": {
      "command": "node",
      "args": ["<path-to-your-index-js-file>"],
      "env": {
        "DOVETAIL_API_TOKEN": "<YOUR_TOKEN>"
      }
    }
  }
}

Restart your client and the Dovetail tools will be available.

Use Cases

Summarize all customer feedback in a project and have the agent draft a PRD grounded in linked highlights and insights.
Pull supporting quotes and existing insights for a hypothesis (e.g. "checkout abandonment due to shipping cost") using search_workspace.
Auto-generate weekly research digests by listing recent insights across multiple projects and rendering them as markdown.
Let a product analyst ask "What are the top complaints about onboarding?" directly in Claude or Cursor without copy-pasting from Dovetail.
Cross-reference interview transcripts with feature plans by fetching get_data_content for specific sessions while coding in Cursor.

Example Prompts

"Search my Dovetail workspace for feedback about pricing in the last quarter and summarize the top three themes."
"List the insights in the Onboarding Redesign project and give me the markdown of the top-rated one."
"Find quotes that support the hypothesis that mobile users struggle with our filter UI."
"Pull the transcript for last week's interview with the Acme Corp design lead and extract action items."
"Draft a one-pager for the checkout team grounded only in highlights and insights from the Checkout project."

Pros

Officially built and maintained by Dovetail, not a community fork.
Two deployment modes: hosted Streamable HTTP with OAuth 2.1 for modern clients and self-hosted STDIO for Claude Desktop.
Respects user-level permissions so agents only see data the connected user already has access to.
Returns insight and transcript content as clean markdown, which works well for LLM grounding.

Limitations

Read-only. The server cannot create, tag, or modify projects, insights, highlights, or data in Dovetail.
Requires a Dovetail account, and most useful research data lives in paid workspace plans.
No native one-click setup for Claude Desktop, which still requires the local Node.js STDIO build.

Alternatives

Notion MCP server if your research repository lives in Notion databases rather than Dovetail.
Grain MCP integrations for teams that primarily work from call recordings and transcripts.
Self-built MCP server against the Dovetail REST API for custom write-enabled workflows.