Linear issues to pull requests with an AI coding agent

How it maps to Linear

Queues are workflow states. init resolves your team’s states once and stores their ids:

Role	Typical state
Ready	Todo, or a triage-complete state
Claimed	In Progress
Review	In Review
Parked	Blocked

Claiming is atomic: claudrunner sets the assignee and the state in one update, so two runs cannot take the same issue.

Priority and sub-issues

• Linear priority is numeric and 1 is urgent — claudrunner sorts it the right way round.
• Sub-issues inherit context from their parent. The agent reads the parent before judging whether the work fits in one safe pass.

Set it up

/plugin marketplace add samiraklf/claudrunner
/plugin install claudrunner
/claudrunner:init

Choose Linear, confirm the state mapping, and add one secret:

Variable	Value
LINEAR_API_KEY	A personal API key from Linear settings, ideally for a runner account

As with every board, the key lives in the orchestrator’s environment — the AI agent never receives it.

What arrives

A pull request with a five-part description, tests that ran, and the review findings listed. The Linear issue moves to In Review with the link. Anything unclear goes to Blocked with one question in a comment, so your team answers it in Linear, where they already are.

Next

• Nightly AI security scan — findings land as Linear issues too.
• How a run works

/plugin marketplace add samiraklf/claudrunner
/plugin install claudrunner
/claudrunner:init