Dovetail

Search a Dovetail workspace and read transcripts, highlights and contacts.

Add to Agentman Studio

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

Resources

What the Dovetail connector does

The Dovetail MCP server connects a Dovetail workspace — a customer-research repository holding interview transcripts, highlights and feedback — to Claude, ChatGPT and any MCP-compatible agent. Dovetail documents 40 tools that search the workspace, read research content and create projects, docs, highlights and comments. Sign-in is OAuth with real read and write scopes.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Dovetail tools (40)

  • get_project_highlights
  • get_project_insight
  • get_insight_content
  • list_project_insights
  • get_data_content
  • get_project_data
  • list_project_data
  • get_dovetail_projects
  • search_workspace

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • Anthropic's directory entry is badly out of date. It lists 9 tools against Dovetail's documented 40, omits all 8 write tools, records the permission as Read, and still carries three deprecated insight names Dovetail's docs no longer use. Treat the directory listing as a historical snapshot of this connector, not a description of it.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail or readOnlyHint value is published here. The tool names and groupings come from Dovetail's documentation.
  • Dovetail's own two documents disagree about one scope. Its Claude help page lists contact:write among the scopes granted to Claude; the server's RFC 9728 descriptor does not advertise that scope at all. Read against the descriptor, the connector cannot create contacts, and no create_contact tool appears in the MCP tool tables either. We could not resolve which document is stale.
  • No DCR or CIMD support. Clients requiring dynamic client registration must fall back to a Dovetail API token in an Authorization header. Dovetail exposes no OAuth client credentials to third parties.
  • Rate limits are per workspace, not per user — 200 requests per minute shared across every token in the workspace, so an agent competes with your other integrations. Exceeding it returns HTTP 429 with Retry-After.
  • Metadata reads return no content. List and get calls deliberately omit body text, so an agent must make a second call per record to read anything. On a large project that multiplies requests against the shared rate limit.
  • Presigned download URLs bypass authentication for an hour. download_file returns a URL that anyone holding it can fetch. Do not let an agent paste one anywhere it will persist.
  • The self-hosted server is not the same product. It exposes 8 read-only tools against this connector's 40, and its last release predates most of them. Choosing it for Claude Desktop means choosing a much smaller surface.
  • Whether an agent's comment notifies your teammates is unresolved. Dovetail documents in-product notification settings and a mention feature, but we found no statement that a comment created through the API triggers a notification. Treat it as unknown rather than as a no.

Frequently asked questions

How many tools does the Dovetail MCP server have?

Dovetail's developer documentation lists 40, in five groups. Anthropic's directory lists 9. The 31 the directory omits include every one of Dovetail's 8 create tools, so a reader who trusts the directory entry sees a read-only connector and gets a connector that writes. Build your expectations from Dovetail's list, not Anthropic's.

Can the Dovetail connector write to my workspace, or only read?

It writes. Dovetail documents 8 create tools covering projects, folders, docs, data entries, tags, transcript highlights, doc comments and channel data points. Anthropic's directory entry lists none of them and records the permission as Read, which is wrong. The OAuth scopes Dovetail grants Claude include project:write, note:write, insight:write and channel:write.

What OAuth permissions does the Dovetail MCP server request?

Thirteen application scopes that genuinely separate reading from writing. The server's RFC 9728 descriptor advertises project, note, insight and channel in both read and write forms, plus read-only contact, field, file, search and user. Unlike most connectors, declining a write scope is a real boundary here rather than a formality.

Can the Dovetail connector delete anything?

No. Dovetail's MCP tool list contains no delete, remove or archive verb, and no scope ends in delete even though the underlying REST API documents delete endpoints for projects, docs, data, channels, contacts and comments. The MCP surface is a deliberate subset. Every write it can perform adds a record rather than removing one.

Can the Dovetail connector post a comment that my teammates see?

Yes. create_comment posts to a doc's comment thread, which is a shared surface your colleagues read in the Dovetail app. If no thread exists one is created automatically, and the comment is attributed to the token's owner. Dovetail caps comment length at 10,000 characters and allows restoring a deleted comment for 30 days.

What personal data can the Dovetail connector read?

Interview transcripts and the contact records of the people in them. Dovetail's contacts database holds research participants and customers, with custom fields typed EMAIL and PHONE. Data entries hold transcripts, survey responses and support tickets. list_users returns workspace members. All of it is third-party personal data by construction.

Does the Dovetail connector work with Claude Code?

Yes, and it inherits the connector once you add it to your Claude account. Dovetail states the connector is automatically available in Claude Code and that /mcp confirms it. To add the server directly, run claude mcp add --transport http dovetail https://dovetail.com/api/mcp. Claude Desktop needs the separate self-hosted STDIO server instead.

Are there rate limits on the Dovetail MCP server?

Yes, 200 requests per minute per workspace. Dovetail documents that the limit is shared across all tokens in a workspace, so multiple integrations compete for the same budget. Every response carries X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Reset, and exceeding it returns HTTP 429 with a Retry-After header.

Sources

More Productivity connectors

  • 6sense

    Ask an AI agent about 6sense account insights, buying signals and keyword trends.

  • Adobe Workfront

    Find, create, update and delete Workfront work items from an AI agent.

Browse all MCP connectorsClaude and ChatGPT setup guideBrowse AI skills

Product names and logos belong to their owners. Agentman is not affiliated with a connector's maker unless the page says the connector is built and run by Agentman.

Use it in an agent

Put Dovetail to work.

Connect Dovetail once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio