Linear MCP setup: read issues and limit access

Choose an unimportant issue that you can open yourself. Record its identifier, team, title and a short passage. Keep this note unchanged during setup so you have an independent expectation for the response. The steps here propose checks for your own environment; they are not evidence of an authenticated test against a Linear account.

Choose read access deliberately

Linear documents https://mcp.linear.app/mcp as its Streamable HTTP endpoint with read and write access. Interactive setup uses OAuth. It offers https://mcp.linear.app/mcp/readonly for read tools only; alternatively, the standard endpoint can use the read OAuth scope. Choose the read route for reporting and inspect the functions offered after authorization.

A Cursor remote connection example

Add this entry to the Cursor MCP configuration. Extend an existing file inside mcpServers. The example uses the documented read endpoint and contains no key. Complete the sign-in and check the chosen account and workspace in the browser. Multiple signed-in accounts can otherwise lead to an unexpected data scope. The Cursor guide explains which configuration file to choose.

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

A client with native remote support does not need an additional local launch command for this route. If your client supports only local processes, consult Linear’s proxy instructions separately. Do not combine URL configuration and process arguments in one entry. The client directory covers other configuration formats. Inspect the actual tool list again after converting a configuration.

Match the issue identifier and contents

1. Inspect the read tools offered after authorization. Record the connection and selected account without copying credentials into the note.

2. Retrieve the prepared issue by its identifier or an equivalent unique reference. A title-only search may return several matches.

3. Compare the identifier, team and known passage with the interface. Check whether the result contains the full content or only a summary.

4. Only then expand the test to several issues. Record the filters and response boundaries before turning the result into a team report.

Investigate empty results and later writes

If the known issue is missing, check the account, workspace and team first. Try the same reference in Linear itself, then repeat a small request without additional search filters. An empty response proves neither that the issue was deleted nor that the entire connection is broken. For failed authentication, use the troubleshooting navigator to isolate the failure before investigating content retrieval.

For later write tasks, define a separate test using a test issue and a predetermined field. Read every change back and inspect the destination before retrying. A lost response does not establish that the change failed to arrive. The read-only guide explains the different access layers. Find other implementations through the Linear registry search and compare their declared origin and connection method.

Common Linear setup questions

Which endpoint suits a report that only reads?

Linear documents /mcp/readonly for read tools. Also check the selected account and data scope.

Is an API key mandatory?

The interactive route described here uses OAuth. Never enter credentials in public search fields.

Why is a matching title insufficient?

Titles can occur more than once. Compare the identifier, team and contents of the intended issue.

Does authorization prove synchronization works?

No. Check retrieval, field mapping, changes and retries separately for your actual workflow.

Provider source checked on 2 October 2026. The test steps are editorial suggestions.

1. Linear: MCP endpoints and authentication
   Show retrieval command

   curl -s https://linear.app/docs/mcp

All posts