Linear MCP Server

What it does (plain English)

Linear is an issue tracker many product teams use to plan and ship work. Its MCP server (a plug-in that gives Claude a new ability) lets Claude find, create and update Linear objects, including issues, projects and comments. Linear hosts it, so there’s nothing to install, and it signs you in with OAuth 2.1.

There are two addresses, and the difference matters:

URL	Access
https://mcp.linear.app/mcp	Read and write (default)
https://mcp.linear.app/mcp/readonly	Read only
https://mcp.linear.app/sse	Deprecated fallback for clients without Streamable HTTP

Which one to use? If you only want Claude to answer questions about your backlog, the read-only URL removes any chance of it creating or editing issues by mistake. Switch to the default URL once you’re happy with how Claude handles your workspace. You can add both under different names (for example linear-read and linear-server) and enable only the one you need.

Is it for you?

Pick it if your team (or just you) tracks work in Linear and you spend time copying notes into issues by hand. It’s useful for founders who plan in documents and want that plan turned into a backlog, and for anyone who wants a quick “what’s the state of this cycle?” without opening the app.

If you don’t use Linear, this does nothing for you. GitHub Issues users should look at the GitHub MCP server instead.

What you can do with it

Linear’s docs list these as example workflows:

• Roadmap from a document: “Here’s my product plan. Create a Linear project for it with one issue per feature.”
• Standup notes to comments: “Post each of these standup updates as a comment on the matching issue.”
• Investigate a bug: “Read issue ENG-142 and its comments, then list what we still don’t know.”
• Cycle summary: “Summarise what the Web team finished and what slipped in the current cycle.”
• Implementation plan: “Break ENG-150 into sub-issues with a short plan for each.”

How to set it up

1. claude.ai on a Team or Enterprise plan. Go to https://claude.ai/customize/connectors and connect Linear.

2. Claude Desktop on Free or Pro. Open Settings > Connectors and add the Linear connector. Sign in to Linear when prompted.

3. Claude Code. In your terminal:

   claude mcp add --transport http linear-server https://mcp.linear.app/mcp

   Then start a Claude Code session and run /mcp to complete authentication.

4. Read-only instead. Use https://mcp.linear.app/mcp/readonly in the command above, or request only the read OAuth scope.

5. Using an API key. If you’d rather not use OAuth, send Authorization: Bearer <token> with a Linear API key. For read-only API access, create a key with only the Read permission:

   claude mcp add --transport http linear-server https://mcp.linear.app/mcp --header "Authorization: Bearer YOUR_LINEAR_API_KEY"

   (The --header flag is Claude Code’s standard way to send a header; Linear’s docs describe the header, not this exact command.)

Watch out for

• One workspace per sign-in. Linear says each workspace needs its own auth context, and reconnecting alone does not switch workspaces.
• Writes are on by default. The default URL can create and change issues. If you’re experimenting, start with the read-only URL.
• “Internal server error” on connect. Linear suggests deleting ~/.mcp-auth and retrying, and updating Node.js if needed.
• WSL users can fall back to the SSE URL if the default one fails.
• Okta-managed sign-in. If your company uses Okta with SAML, Linear supports enterprise managed authentication so Claude signs users in through Okta policies.
• Tool list isn’t published in full. Linear’s docs describe the capabilities and say more is coming, so check what appears under /mcp in your client.

Alternatives

• GitHub MCP if your issues live in GitHub.
• Atlassian Rovo MCP for Jira.
• Notion MCP if your tasks are a Notion database.