A task queue is the reviewable plan Coder builds from your mission. It is the same in every execution mode and on every runner.
How a queue is built
Planning expands the mission into tasks across three phases:
- Recon*: read the relevant code, docs, and dependencies to understand the current state.
- Decomposition*: break the objective into ordered, verifiable tasks.
- Risk*: flag risky or external steps that need an approval gate before they run.
Why a queue helps
- You see the work before it happens.
- Each task has a clear success check.
- Risky steps are surfaced for approval instead of running silently.
- The same queue drives local, cloud, and hybrid runs.
Reviewing a queue
Before execution, review the proposed tasks. You can adjust scope, drop tasks, or require approvals on specific steps. Once approved, Coder executes the queue in order.
Related
- Tasks run through [agents](/docs/coder-agents) in isolated workspaces.
- Risky steps go through [approvals](/docs/coder-approvals).
Was this article helpful?
