The Triage Queue — inbox captures, queue clears.
The inbox captures everything. The queue is where you work through it — surfacing unreviewed items, priority bugs, and agent assignments in a structured view. A coding agent can work through the queue directly via the API, or hand off anything that needs human judgement.
Inbox vs Queue
Two surfaces, one pipeline.
Inbox and Queue are distinct surfaces in the Quell admin. Inbox is the raw stream — everything that arrived, ordered by time. Queue is structured triage: filtered, prioritised, and assignable. An agent working the queue sees only what it can act on, not the full stream.
Inbox
Every submission, in order. Raw and unfiltered. Review what came in, pick what matters, set status and assignee.
Queue
Structured triage. Unreviewed items surfaced first. Priority ordering. Agent assignments tracked. The place you clear the backlog.
How the queue works
Surface, triage, automate.
Agent-native triage
An agent that clears the queue.
Give a coding agent a scoped API key and it can list open items, read each report's full context, update status, comment its reasoning, and open issues in GitHub or Linear. Trust-aware routing controls which submissions the agent can act on autonomously. Anything that needs human judgement gets a comment and stays open.
- List open items.
GET /api/feedbackwith status and priority filters returns the queue in the agent's working window. - Read context.
GET /api/feedback/:id/contextreturns Markdown — console trace, network failure, screenshot, page state — in one call. - Act. Comment reasoning, update status to in-progress or resolved, open a GitHub/Linear issue if the report warrants it.
- Hand off. Ambiguous items get a comment explaining why they need human review. The team picks them up from there.
FAQ
Queue questions.
What is the difference between the inbox and the queue?
The inbox is where all submissions arrive — raw, unreviewed, ordered by time. The queue is a structured triage surface: it surfaces what's unreviewed, what's assigned, and what's been acted on, with priority ordering and status filters. Inbox captures; queue clears.
Can an agent work through the queue automatically?
Yes. An agent with a <code>feedback:read</code> + <code>feedback:write</code> + <code>feedback:integrations</code> scoped key can list open items, read each one's <code>/context</code> Markdown export, update status, comment, and open GitHub or Linear issues. Trust-aware routing determines which submissions the agent can act on without human review.
How does priority ordering work?
Each feedback item carries a 2×2 priority field: the reporter sets an urgency signal (low / high) and type (bug / idea / praise) at submission time. The queue surfaces high-priority bugs first. Agents can read and act on the priority field directly.
What happens to items the agent can't resolve?
Items that require human judgement — ambiguous reports, policy decisions, trust-level escalations — stay in the queue with a comment from the agent explaining why it was left open. The team sees the agent's note alongside the full context and can act from there.
Is the queue different from a Linear or Jira board?
Yes. Linear and Jira are issue trackers — destinations for triaged and confirmed work. The Quell queue is upstream: it's where raw feedback is reviewed, structured, and decided on. Quell creates the issue in Linear or GitHub when the decision is made; it doesn't replace those tools.