AgentQueue syncs tickets, subtasks, and blockers into a live execution graph. Then the agent runs independent tickets in parallel and sequences dependent work automatically, opening small PRs you can review and merge.
Capabilities
Queue-first board for backlog → planned → coding → PR ready. Drag tickets in, reorder priorities anytime, and let the agent schedule execution based on what’s blocked vs unblocked.
Every ticket starts with a short implementation plan in GitHub (or as a Linear comment). Approve, edit, or send feedback—then the agent writes code.
When tickets depend on each other, the agent generates a clean chain of stacked PRs—each small, reviewable, and based on the previous. Automatic rebasing keeps the chain clean.
Plan approvals, PR review requests, CI failures, and agent questions in one place—so nothing gets lost and you can unblock work fast.
Works with your existing branch rules, CODEOWNERS, required checks, and merge policies. The agent follows your naming conventions and never bypasses CI.
Protected paths, diff-size limits, and mandatory checkpoints. The agent won’t merge without passing required checks and getting the approvals you already enforce.
Workflow
Connect your Linear workspace and GitHub repos. AgentQueue maps tickets → branches/PRs and follows your repo rules (checks, approvals, naming).
Drag tickets into the queue and set priority. AgentQueue reads subtasks and blockers to build a live execution graph—then schedules parallel work automatically.
The agent drafts a plan first. Approve, edit, or send feedback—then the agent starts coding. No guessing, no surprise diffs.
The agent opens a PR, runs CI, and requests review. You merge—and the agent continuously advances the graph as tickets become unblocked.