Linear MCP Server: Setup and Security

The Linear MCP server is Linear's own hosted server at https://mcp.linear.app/mcp. It lets an agent read and manage issues, projects, and cycles. It runs over HTTP with OAuth 2.1 — scoped, revocable tokens, no API key in a config file. TrustedMCP grades it A; it is one of the better-secured integrations in the directory.

What it connects

• Read: issues, projects, cycles, comments, teams, and users in the workspaces you authorise.
• Write: create and update issues, move them through states, add comments — bounded by the OAuth scopes you grant at connect time.
• Runs remotely, so there is no local process and no stdio command-execution surface.

Setup

Claude Code:

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

Cursor — ~/.cursor/mcp.json:

{ "mcpServers": { "linear": { "url": "https://mcp.linear.app/mcp" } } }

VS Code — .vscode/mcp.json:

{ "servers": { "linear": { "type": "http", "url": "https://mcp.linear.app/mcp" } } }

Claude Desktop uses the same URL via a custom connector. On first use the client opens Linear's OAuth screen; approve the workspace and scopes there. Nothing secret is stored in the config file — the token lives in the client's credential store and can be revoked from Linear's settings.

Security notes

The remote OAuth model removes the two biggest MCP risk classes at once: no long-lived key sits in a checked-in file, and the HTTP transport avoids the stdio command-execution issue entirely. Two things still apply:

• Grant the narrowest scope. If the agent only triages, it does not need issue-delete or admin scope.
• Issue and comment text is untrusted input. A ticket body can carry instructions aimed at the model, so keep any high-impact follow-on action (closing a sprint, bulk edits) behind a human confirmation. See MCP security best practices.

Full trust breakdown on the Linear MCP listing; the scoring method is in how TrustedMCP scores servers. For enabling it in the editor, see the VS Code MCP guide.