Linear MCP Server

The Linear MCP Server is the official MCP entry point to Linear, available since 1 May 2025. It is centrally operated and hosted by Linear — there is nothing to install locally; the client connects directly to the remote endpoint. According to the provider, the server follows the authenticated remote MCP specification and uses Streamable HTTP as its primary transport. No public GitHub repository is available for this server; the repository linear/linear-mcp-server does not exist on GitHub.

What is the Linear MCP Server

The Linear MCP Server gives MCP-capable AI clients such as Claude, Claude Code, Cursor, Codex, VS Code, Windsurf, or Zed structured access to your own Linear workspace. According to the provider, the server exposes tools for finding, creating, and updating objects in Linear — issues, projects, and comments. Per Linear, more functionality is being added over time.

Typical tasks an agent can handle with the Linear MCP Server: automatically creating issues from a meeting note, updating an existing issue's status and assignee, summarizing a cycle's open tasks before a standup, or structuring an implementation plan directly as tickets.

Endpoints and Transport

The default read-write endpoint is https://mcp.linear.app/mcp using Streamable HTTP transport. For read-only usage — such as automated status reports — the variant https://mcp.linear.app/mcp/readonly is available, exposing only read tools. A further option is to request the standard endpoint with the read OAuth scope: clients that only request read access receive a token that cannot reach write APIs.

The older SSE endpoint https://mcp.linear.app/sse is, according to the provider, a deprecated fallback for clients without Streamable HTTP support and should not be used for new setups.

Authentication

Sign-in uses OAuth 2.1 with dynamic client registration, per the provider: on the first connection, the Linear login opens and you confirm access for the client. A bearer token in the Authorization header and a Linear API key are also supported. Organizations using Okta can manage access via Okta SAML: after SAML setup via Okta's OIN, MCP enterprise authentication can be enabled from the Okta identity provider section in Linear settings. The issuer URI is then pasted into Linear settings, allowing compatible clients like Claude to authenticate users automatically via Okta access policies.

Setup by Client

According to the official documentation, the Linear MCP Server supports a wide range of clients directly:

Claude Code — Set up with a single command: claude mcp add --transport http linear-server https://mcp.linear.app/mcp. Then run /mcp in a Claude Code session to start the authentication flow.

Claude Desktop / Claude.ai — Via Settings → Connectors → add the Linear connector. Team and Enterprise users connect via the Connectors page.

Cursor — Directly through Cursor's MCP tools page or by clicking the provided installation link.

VS Code — Via CTRL/CMD+P → "MCP: Add Server" → Command (stdio) → npx mcp-remote https://mcp.linear.app/mcp.

Windsurf — Via CTRL/CMD+, → Cascade → MCP Servers → Add Server → JSON configuration with npx -y mcp-remote https://mcp.linear.app/mcp.

Zed — Via CMD+, → add a context_servers block in Zed settings.

Codex — codex mcp add linear --url https://mcp.linear.app/mcp in the CLI; for first use, the RMCP feature must be enabled in ~/.codex/config.toml.

Requirements and Cost

The MCP server itself is provided without a separate fee, according to the provider. A Linear account is required; Linear offers a free plan as well as paid tiers such as Basic, Business, and Enterprise. The current plan overview is available on the provider's official pricing page. Through the server, the agent only sees the data the signed-in Linear account already has access to — visibility rules, team memberships, and permissions remain in effect unchanged.

Why No Local Install

Unlike many MCP servers for developer tools, Linear deliberately does not offer a locally installable option. Since the server operates purely on the permissions of the signed-in Linear account and needs no logic outside the Linear API, the usual benefits of running it locally largely disappear. Updates, security patches, and new features arrive automatically through the central endpoint, without users having to keep their own version current.

Security and Data Access

Since the server exposes write tools, it should only be set up in trusted client environments. The connected AI agent can create and edit issues, projects, and comments without additional confirmation — provided the signed-in Linear account has the corresponding permissions. For scenarios where only reading is needed, the readonly endpoint is recommended to structurally rule out accidental write access. Prompt injection via Linear content is a general risk when AI agents process external data.

Frequently Asked Questions

Can I connect multiple Linear workspaces at once? Yes, it is enough to register the server multiple times under different names in the client and sign in to each workspace when connecting.

Does the agent also see private teams? Only if the signed-in account itself has access to that team; Linear's visibility rules remain in effect unchanged.

Does the server work with the free Linear plan? Yes, MCP access itself is, according to the provider, usable regardless of the chosen plan.

Is there a public GitHub repository? No. The repository linear/linear-mcp-server does not exist on GitHub (404). The Linear MCP Server is a proprietary hosted service with no published source code.

Active Linear account and an MCP-capable AI client supporting Streamable HTTP, or the mcp-remote npx module for older clients.

For Claude Code: claude mcp add --transport http linear-server https://mcp.linear.app/mcp, then run /mcp in the session. Add other clients via JSON configuration with npx -y mcp-remote https://mcp.linear.app/mcp. Complete the OAuth login on first connection.

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

OAuth 2.1 with dynamic client registration. Alternatively: bearer token or Linear API key. Enterprise: Okta SAML integration available.

Read and write access to issues, projects, and comments, limited to the permissions of the signed-in Linear account. Read-only endpoint available for read-only usage.

The server returns tool results; the connected MCP client decides which data is forwarded to the AI model.

Write tools can modify issues and projects without confirmation. Use the readonly endpoint for read-only scenarios. Prompt injection via Linear content is a general risk when using MCP servers with write access.