Linear MCP server: connect AI agents to Linear from the EU

Linear is an issue tracker and project planning tool used by many software teams, made by Linear Orbit, Inc. in San Francisco. Teams use it for issues, cycles, projects, roadmaps and product documents.

Linear's official MCP server runs at https://mcp.linear.app/mcp over Streamable HTTP. It authenticates with OAuth 2.1, including dynamic client registration, or with a Linear API key sent as a Bearer token. A read-only endpoint, https://mcp.linear.app/mcp/readonly, exposes only the reading tools, which suits agents that should never change your workspace.

A triage agent can read new bug reports from a support channel, check Linear for duplicates and create a well-formed issue with steps, labels and priority, linking the original conversation. A planning agent can summarise what shipped in the last cycle and draft release notes from completed issues.

Engineering leads can ask "which issues in this project are blocked and why?" and get an answer with links. Agents can also keep Linear in sync with GitHub by commenting on issues when related pull requests change.

Add the Linear connector on inference.eu and sign in with Linear through OAuth. You choose the workspace, and the agent acts with the permissions of the user who authorised it. Alternatively, paste a personal API key with restricted permissions.

If an agent should only read, point the connector at the read-only endpoint. Tokens and keys are stored encrypted in the EU and are used only by the agents you allow. Workspace admins can review and revoke OAuth applications and API keys in Linear's settings.

Issues, projects and comments are fetched from Linear's own MCP server and API, so data stays with Linear, a US company, as it does for your other Linear integrations.

On inference.eu, the model that reads and writes your issues runs in EU data centres, and your Linear credentials are stored encrypted in the EU. Issue descriptions often contain customer names, error logs and internal plans, so connect agents only to the teams they need and prefer the read-only endpoint for reporting and research agents.

Creating and editing issues is reversible, but a busy agent can flood a backlog, reassign work or change priorities in ways that disrupt a team. Comments are visible to everyone in the team and may notify people.

On inference.eu you can require a person's approval for creating issues, changing status, priority or assignee, and posting comments, while letting searches and reads run on their own. Starting with the read-only endpoint and adding write access once an agent has proved itself is a sensible path.