# Linear MCP

> Linear's hosted MCP server — find, create, and update issues, projects, and comments from your AI agent, with OAuth and one-command setup.

Linear's centrally hosted MCP server connects agents to your issue tracker: find, create, and update issues, projects, and comments. One command (claude mcp add --transport http linear-server https://mcp.linear.app/mcp), an OAuth browser flow, and your agent can turn 'fix LIN-123' into reading the actual ticket — and closing it with a comment when the PR is up.

Website: https://linear.app/docs/mcp

Linear MCP makes the issue tracker part of the agent's working memory. Instead of copy-pasting ticket text into prompts, the agent reads `LIN-123` itself — description, comments, status — implements against it, and writes the follow-up comment when the work is done.

## Highlights

- **The core loop covered** — finding, creating, and updating issues, projects, and comments; the surface keeps growing (product-management tools and agent-support landed through 2026).
- **Hosted and managed** — Linear runs it (built with Cloudflare and Anthropic at launch in May 2025); you configure a URL, not a process.
- **OAuth with dynamic registration** — the `/mcp` browser flow handles auth; headless contexts can pass an `Authorization: Bearer` token instead.
- **Streamable HTTP** — the modern transport at `mcp.linear.app/mcp` (the older `/sse` endpoint survives mainly for WSL compatibility).

## In an AI-assisted workflow

```bash
claude mcp add --transport http linear-server https://mcp.linear.app/mcp
# in a session: /mcp → linear-server → Authenticate
# then:
# > Read LIN-482, implement the fix, and comment with a summary + the PR link
```

The pattern that pays: let the ticket be the spec. The agent pulls acceptance criteria from the issue rather than your paraphrase, and status updates land where the team already looks.

> [!TIP]
> Pair it with the [GitHub MCP server](/tools/github-mcp-server) and the loop closes end-to-end: Linear holds the *why*, GitHub holds the *what*, and the agent moves both — issue → branch → PR → comment — without you ferrying context between tabs.

## Good to know

Free with your Linear workspace, Web-only (it's a hosted remote), no public repo — Linear operates it centrally and ships updates via their changelog. WSL users who hit connection errors on the HTTP endpoint have a documented workaround through `mcp-remote` against the legacy SSE endpoint. As with any write-capable server, consider an `ask` rule on issue-mutating tools in [your permissions](/guides/configuration/claude-code-settings-permissions) until you trust the loop.

---

_Source: https://agentscamp.com/tools/linear-mcp — Tool on AgentsCamp._