> mcp integration
Linear MCP Server: What It Does & How to Use It with Claude
What the Linear MCP server can do — find, create, and update issues, projects, and comments — and how to connect it to Claude Cowork, Code, or Desktop.
The Linear MCP server lets an AI assistant read and act on your Linear workspace, finding, creating, and updating issues, projects, and comments through natural language. Linear runs it as a remote, OAuth-authenticated server at https://mcp.linear.app/mcp, so you connect by URL with no local install.
What the Linear MCP server can do
The official server exposes tools for finding, creating, and updating the core Linear objects. Per Linear’s docs, it covers issues, projects, and comments, with more functionality rolling out over time.
| Capability | What it enables |
|---|---|
| Find issues | Search and filter issues by team, status, assignee, label, or free text, useful for triage and standup summaries |
| Create issues | Turn a sentence or a chat thread into a properly-fielded issue (title, description, team, priority) |
| Update issues | Change status, priority, assignee, labels, or any field on an existing issue |
| Projects | Look up and update projects so an assistant can place work in the right initiative |
| Comments | Read and add comments on issues to capture context or post an AI-drafted update |
If a capability is not in that list (bulk imports, deep analytics, workflow automation rules), the official server does not promise it today. Linear explicitly notes more is “on the way.”
Official vs community server
- Official (
mcp.linear.app/mcp) is the right default. It is centrally hosted and managed by Linear, uses OAuth 2.1 with dynamic client registration, and stays scoped to issues, projects, and comments. Nothing runs on your machine, and Anthropic clients connect to it natively. - Community (
@tacticlaunch/mcp-linear) is a local stdio server (Node, run vianpx) authenticated with a personal Linear API key. It reaches further: milestones, roadmaps, cycles, custom fields, documents, and notifications. Use it when you need surface area the official server has not shipped yet, and you accept running and updating it yourself.
How to connect it to Claude (Cowork / Code / Desktop)
- Claude Cowork / Desktop: open Settings → Connectors → add a custom connector and paste
https://mcp.linear.app/mcpas the remote MCP URL. - Claude Code: run
claude mcp add --transport http linear-server https://mcp.linear.app/mcp, then run/mcpin a session to finish auth. - Authenticate: complete the OAuth flow in the browser; you sign into your existing Linear account, so the assistant inherits exactly your permissions.
- Start read-only. Ask it to find and summarize issues before you let it create or update anything, so you can confirm it is hitting the right teams and projects.
How we use it: good for converting messy meeting notes and Slack threads into clean, well-fielded issues, and for fast “what’s blocked in this cycle?” triage. Bad for reporting or bulk edits: it is a per-object tool, not an analytics layer, and it will not bulk-move fifty issues for you. One real gotcha: in a large workspace the model often guesses the wrong team or project on a fresh issue, so always name the team explicitly in your prompt (or have it confirm the target before creating).
Security & permissions
The server authenticates with OAuth 2.1 against your own Linear account, so it can never exceed your permissions. Narrow the account’s access if needed, and start read-only before granting create/update in a shared workspace.
Wiring Linear into your stack
The Linear MCP server is the fast path to a useful workflow, but the value shows up when it is wired alongside your other tools: Slack for intake, GitHub for the code, and clear guardrails on what the agent may write. That orchestration is what we do. Book a discovery call.
Sources: Linear MCP docs · Linear MCP changelog · tacticlaunch/mcp-linear