What Linear MCP Server Does
Linear's official MCP server lets AI clients work with Linear issues, projects, teams, cycles, and workspace search through a hosted remote endpoint at mcp.linear.app/mcp. Linear launched the server on May 1, 2025 and credits Cloudflare plus Anthropic for infrastructure support in the changelog. For product and engineering teams, the value is straightforward: an agent can look up current project state, find issues, create or update work items, and connect code changes to the planning system without relying on brittle browser automation. It turns Linear into a native task and project context source for coding agents, project copilots, and internal automation tools.
Two Ways to Authenticate
Linear supports OAuth 2.1 with dynamic client registration over Streamable HTTP, matching the March 2025 MCP spec. That is the path most modern AI clients want because it avoids manually copying long-lived secrets and lets the client handle a standard authorization flow. For teams evaluating remote MCP seriously, OAuth support is a major signal: it means the integration is built around a first-party, standards-aligned connection model rather than a wrapper around a personal API token. It also gives security teams a cleaner place to reason about revocation, user consent, and client registration.
Linear also documents a simpler Authorization: Bearer header option that can carry either a scoped API key or an OAuth access token. That fallback is useful for internal tools, read-only automations, or environments where a full client registration flow is unnecessary. It should not be treated as a shortcut around governance: API keys still need scoping, storage, rotation, and audit. But the presence of both paths makes the server flexible. A polished AI client can use OAuth, while a small internal integration can start with a scoped token and graduate to a more formal setup later.
Included at Every Pricing Tier
Linear's pricing page lists MCP access as included across Free, Basic, Business, and Enterprise plans. That is unusually generous for a new agent integration because many vendors reserve their most useful AI or integration features for paid or enterprise tiers. A team on the free plan can still connect an MCP-capable client and test whether agent workflows around issue triage, project updates, and cycle status are worth expanding. The free tier's broader product limits still matter, but the MCP connector itself is not the paywall.
Paid tiers add the normal Linear product capabilities: Basic at $10 per user per month, Business at $16 per user per month when billed annually, and Enterprise on custom pricing. The important review point is that those tiers do not appear to gate MCP access itself. A buyer should still choose a Linear plan based on team, workflow, compliance, and support needs, not because the MCP server only turns on at a higher tier. That makes Linear one of the cleaner official MCP evaluations: the technical integration can be tested before a procurement argument about connector pricing begins.
Client Compatibility and Fallback Support
Native support spans major AI development surfaces: Claude as a Team/Enterprise connector, Claude Desktop, Claude Code, Cursor, Codex, VS Code, Jules, Vercel's v0, and Zed. Linear also says hundreds of other tools can connect through the mcp-remote npm fallback for clients that do not yet support remote MCP directly. That mix is strong for developer experience because teams can use Linear context from the tool where work already happens. A coding agent can inspect an issue, update status, or create follow-up work without forcing the developer to switch to the Linear UI for every small action.
The mcp-remote fallback is especially practical during the current transition period in MCP clients. Remote MCP support is improving quickly, but not every client implements the latest transport or auth flow at the same pace. A local bridge process lets teams use Linear with clients that would otherwise be blocked. The trade-off is operational: another local process can fail, need updates, or behave differently under WSL and corporate device controls. For production usage, native client support is preferable; for experimentation and long-tail tools, mcp-remote keeps the door open.
Setup Friction and Troubleshooting
Linear's docs include a public troubleshooting section for common connection failures, including clearing the ~/.mcp-auth cache, updating Node, and changing transport settings under WSL. Those details are not glamorous, but they are a positive signal. Remote MCP depends on a chain of client support, local environment, browser/OAuth flow, token cache, and transport compatibility. Documented failure modes reduce the time a team spends guessing whether a problem is in Linear, the AI client, Node, WSL, or the local bridge.
None of the documented issues look unusual for an OAuth-based remote MCP integration. They are the normal first-connection problems that show up when a standard is still young and client implementations differ. The best rollout pattern is to test with one or two supported clients, document the exact setup path, and avoid mixing native remote MCP and mcp-remote fallback modes in the same pilot unless there is a reason. Once the first working path is stable, Linear's free-tier inclusion makes it easy to expand the test to more users without a connector-specific procurement step.
The Bottom Line
Linear's MCP server is one of the most developer-friendly official integrations in the current MCP category. It is hosted, standards-aligned, free-tier accessible, documented with OAuth and bearer-token options, and supported across a wide set of AI coding clients. The main cautions are ordinary integration hygiene: scope API keys carefully, handle OAuth cache issues, understand workspace permissions, and prefer native client support over a bridge process when possible. If your team already runs planning in Linear, this is a high-signal MCP integration to try early because the pricing gate is low and the workflow value is direct.