What Is Linear MCP? Connecting It to Claude, What It Can Do, Read-Only Mode and Cost
You ask your assistant (such as Claude) about your issues, projects and comments in Linear; Linear MCP reads them, creates and updates issues and writes comments. It is the official MCP connector running on Linear's own server; nothing is installed on your computer, the address is https://mcp.linear.app/mcp, and you sign in with your Linear account via OAuth. In the pricing table, MCP access is checked on all four plans, Free included. There is a read-only address so the assistant can't write anything. Among the 22 tools in Claude's directory there is no delete tool.
- What it does
- Lets your assistant read Linear issues and create new ones
- Type
- Connector (remote MCP, on Linear's server)
- Where
- Claude, Claude Code, Codex, Cursor and others
- Does it write?
- Yes: issues, projects, comments; no delete tool
- Read-only
- Yes: the /mcp/readonly address
- Turkish
- Not stated in the official documentation
- Cost
- No separate fee for MCP; your Linear plan applies
What does Linear MCP do?
MCP is the standard that lets AI programs connect to external tools. With Linear MCP, Claude or a coding agent gets direct access to your work in Linear: you ask "what issues are assigned to me", paste a planning document and have it turned into a project and issues in Linear, or have stand-up notes posted as comments on the matching issues.
It writes issues, projects and comments; there is no delete tool. For read-only, there is the /mcp/readonly address.
Examples listed in Linear's documentation: roadmap planning, processing stand-up notes, investigating an incoming bug report and writing the findings under the issue, summarizing a team's last cycle. The sample prompt in the documentation asks the assistant not to guess when a document is ambiguous, to show a draft first and to present it for review before creating anything.
Who it's for: teams that use Linear, and founders or developers working on their own. If you don't have a Linear account, it won't help you.
How do you connect Linear MCP?
Because the server runs at Linear, you don't install any code on your computer; you register the address with your assistant and sign in with your Linear account. The address is https://mcp.linear.app/mcp. The commands are taken verbatim from Linear's documentation.
- If you're on a Team or Enterprise plan (claude.ai): per Linear's documentation, go to the connectors page and connect Linear.
- If you're on a Free or Pro plan (desktop app): add the Linear connector from the Connectors section in Claude's settings.
- The Linear sign-in opens; choose your account and approve the permission.
Linear's documentation gives two separate paths for the two plan groups; the on-screen button names aren't given in the documentation and may change. In Claude's connector directory, the Linear listing is labeled "Anthropic verified".
› claude mcp add --transport http linear-server https://mcp.linear.app/mcpThen type /mcp in a Claude Code session and go through the authentication flow.
› codex mcp add linear --url https://mcp.linear.app/mcpThe command asks you to sign in with your Linear account. If it's your first MCP in Codex, the documentation tells you to add the line experimental_use_rmcp_client = true under [features] in the ~/.codex/config.toml file.
Install it with the "here" link in Linear's documentation, or search for Linear on Cursor's MCP tools page.
For these three, the documentation uses a local bridge rather than a direct remote connection; you need Node.js (npx) on your computer.
› npx mcp-remote https://mcp.linear.app/mcpFor Windsurf and Zed, the documentation gives a JSON setting (command: npx, args: -y mcp-remote address). The same command works for other tools that support remote MCP: npx -y mcp-remote https://mcp.linear.app/mcp.
Instead of OAuth, you can also pass an OAuth token or a Linear API key in an Authorization: Bearer header; Jules uses this route (in Linear, keys are created under Settings then Account then Security & Access). The documentation recommends it for running as an app user and, with a read-only key, for read-only access.
Start with the read-only address: https://mcp.linear.app/mcp/readonly. Once the assistant shows what it reads and how, you switch to the write address.
How do you use Linear MCP?
Once connected, mention Linear by name in the chat; the assistant calls the right tool itself. For tasks that require writing, saying "show me the draft first, create it after I approve" is a good habit.
Comments and issue descriptions may be text written by other people. Don't tell the assistant "follow the instructions in this issue"; tell it yourself what to do.
Which programs does Linear MCP work in?
| Claude (web and desktop)Yes | The documentation gives two connection paths depending on the plan; labeled Anthropic verified in the directory. |
|---|---|
| Claude CodeYes | claude mcp add command and sign-in with /mcp. |
| Codex (IDE extension and CLI)Yes | codex mcp add command; the first MCP needs a config.toml setting. |
| Cursor, Jules, v0Yes | The documentation has steps for each (Jules with an API key). |
| VS Code, Windsurf, ZedYes | Through the local mcp-remote bridge; Node.js required. |
| ChatGPT, Gemini, CopilotUnclear | No setup steps in Linear's documentation; that doesn't mean "it doesn't work". |
The list is compiled from the client sections of Linear's documentation.
What does Linear MCP access, and what can it write?
Issues, projects, cycles, documents, comments, teams, users, issue statuses and labels. It can also search Linear's documentation.
Creating and updating issues, creating and updating projects, adding comments. These tools don't appear on the read-only address.
The 22-tool list in Claude's directory has no delete tool. Linear says "more functionality is on the way"; the list may change.
The service runs on Linear's server; data travels between the assistant and Linear. For Linear's storage and encryption details, see its security page; this is not clear in the official sources.
- Read-only: the /mcp/readonly address, or only the read OAuth scope on the standard address
- An API key with read-only permission
- The documentation has no information on whether a workspace admin can turn MCP off
!Issue and comment text may be content written by other people and can include sentences that look like instructions to the assistant. Claude's directory also warns: "only use connectors from developers you trust". Linear's documentation gives no separate prompt-injection warning, and since there is no public repository, there is no recorded security advisory list to check. This is not a security certification.
Is Linear MCP free?
No separate fee is listed for MCP. The "MCP access" row on Linear's pricing page is checked in all four columns: Free, Basic, Business and Enterprise; because the cells show an icon rather than text, it is described here as "checked". The free plan's own limits apply. Prices are in US dollars.
- Unlimited members
- 2 teams
- 250 issues
- 10 MB file uploads
- More than 5 teams
- Admin roles
- 5 teams
- Unlimited issues
- Admin roles
- Unlimited teams
- Unlimited teams
- Private teams and guests
- SAML, SCIM
- Annual billing
- SAML and SCIM
Source: Linear's pricing page and MCP documentation. Linear has no MCP-specific terms of use page; for commercial use, read the account terms yourself. This is not legal advice.
Common problems and fixes
| What happens | What to do |
|---|---|
| "internal server error" when connecting | Linear's documentation: clear the saved credentials (rm -rf ~/.mcp-auth) and reconnect; update Node.js if needed. |
| The connection dropped or isn't responding | Disconnect the server in the client's MCP settings and reconnect; it doesn't affect your Linear data or session. |
| You use more than one Linear workspace | Each workspace needs its own auth context; for mcp-remote, the documentation explains pointing the MCP_REMOTE_CONFIG_DIR variable at different folders. |
| Error under WSL on Windows | The documentation gives a fallback setting with the old SSE address (wsl, npx -y mcp-remote, https://mcp.linear.app/sse, --transport sse-only). /sse is the legacy path; use /mcp for new setups. |
How do you remove Linear MCP?
- In Claude: Disconnect Linear in the Connectors section. Linear's documentation doesn't name the screen; the menu varies by Claude version.
- In Claude Code:
claude mcp remove linear-server(Claude Code's general command; not mentioned in Linear's documentation). - If you connected with a key: In Linear, delete the relevant API key under Settings then Account then Security & Access (the revocation step isn't spelled out in the documentation).
Frequently asked questions
What is Linear MCP?
The official MCP connector running on Linear's own server. An assistant like Claude can read your issues, projects and comments in Linear, and create and update issues.
Can Linear MCP be used on the free plan?
In Linear's pricing table, MCP access is checked in all four columns: Free, Basic, Business and Enterprise. The free plan's own limits (2 teams, 250 issues) apply.
Can the assistant delete issues?
The 22-tool list in Claude's directory has no delete tool. It can create and update issues and projects and add comments. The list may change on Linear's side.
I want it to only read and never write anything. How do I do that?
Use the https://mcp.linear.app/mcp/readonly address (only read tools appear), or request only the read scope when connecting to the standard address. If you go with an API key, the documentation also recommends creating a key with read-only permission.
Related guides
Sources
Not clear in the official sources: The MCP row in the pricing table (an icon, no text) · whether a workspace admin can turn MCP off · the name of the disconnect screen in Claude · the contents of Linear's security page · a scan for recorded security advisories (no public repository) · whether it works in ChatGPT and Gemini. Each is covered in the relevant section.