Aviator Inbox
Know which PR to review next.
A smarter PR inbox that answers three questions on every row, without clicking: should I review this now, how long will it take, and what is it blocking?
Free for individual developers. Works against any GitHub repo.
From startups to Fortune 500s, the world's most effective engineering teams supercharge their developers with Aviator.
The problem
One long list. Every PR looks the same.
A typical review queue blurs together “things I need to act on” with “things waiting on others.” Reviewers scan every row to figure out what to review, in what order, and what it’ll cost.
Question 1
Should I review this now, later, or never?
Inbox groups PRs by what to do next — re-review, high leverage, required, optional, or stale.
Question 2
How much will this cost me?
Every row carries an effort badge tailored to you: ☕ quick, ~10 min, ~30 min, ~1 hr, or deep dive.
Question 3
What's actually blocking it?
A one-line blocking reason on every row — stacked dependents, hotfix label, codeowner requirement, CI.
Four pages, not one
A clean split between action queue and watch list
Inbox is what you need to review. Outbox is your PRs waiting on others. Two more pages — Custom sections and Snoozed — handle bespoke views and PRs you've hidden for later.
- Inbox: PRs you need to review or re-review
- Outbox: your PRs waiting on review, CI, or queue
- Custom sections: user-defined filtered views
- Snoozed: hidden until later
Smart groups
Five groups. Every PR lands in exactly one.
Inbox auto-classifies each PR into one of five groups by precedence. Empty groups auto-hide so your queue stays focused on what actually needs you.
Needs Re-Review
You already approved this PR. The author pushed new commits. We surface only the delta so you don't re-read everything.
High Leverage
Hotfixes, PRs blocking stacked dependents, anything marked Urgent. The work that unblocks other people if you pick it up next.
Required of You
You're a required reviewer or codeowner. The default home for PRs you can't ignore.
Optional / FYI
Others have already approved. You're a watcher, not a blocker. Skim, learn, or skip.
Stale
Idle, low-priority, weekly-triage candidates. Auto-collapses so they don't crowd the active queue.
Per-row signals
Three signals on every row. No clicks required.
Effort, blocking reason, and a plain-language why. Every score and group placement is one sentence away from an explanation — confidently wrong is worse than absent.
- Effort badge tailored to your familiarity, not raw LOC
- Blocking reason that names the dependency, label, or owner
- Plain-language why on every group placement
- Pin to top, Mark Urgent, or Snooze with one click
Scoring
Effort and blocking, computed from real signal
Effort blends lines of code, repetition, file type, and your familiarity with the files. Blocking blends stacked dependents, age, queue intent, manual urgency, and release labels. Both are computed asynchronously, cached, and displayed in wide buckets — trust comes before precision.
- Effort: LOC + repetition + file type + your familiarity
- Blocking: stacked dependents + age + queue intent + labels
- Wide buckets, not precise numbers — trust comes first
- Calibrated against your team's actual review time
All features
Built for the way you actually triage PRs
Surface Inbox where you already work — in the dashboard, in your Slack home page, or in the Chrome extension on github.com.
- Custom sections
- AttentionSet Chrome extension
- Slack home page
- Attention reasons
- Manually change attention
- Repository filtering
- Real-time updates
- Snooze a PR
- Pin to top of group
- Mark Urgent (promotes to High Leverage)
- V1 dashboard fallback during ramp-up
- Best practices for review queues
Roadmap
Five good signals beat twelve great-on-paper ones
Phase 1 ships only what we can verify works. LLM-driven signals run in shadow mode until they meet a precision bar.
- 01
Foundation
4-page split, 5 groups, effort + blocking scores, pin/urgent. No LLM — only existing signal.
- 02
First LLM signals
One-line PR summaries, PR-type classifier, Quick Wins + Routine + Stacked groups.
- 03
Risk + linked issues
High Risk group (shadow mode first), cohesion check, Jira/Linear linked-issue priority.
- 04
Magic features
Re-review delta summary, urgency extraction, thread-resolution status.
- 05
Calibration
Real review-time data calibrates the scores. Advanced complexity metrics.
Stop scanning every PR.
Free for individual developers. Set up in one click against any GitHub repo.
Join us at The Hangar
A vetted community for developer-experience (DX) enthusiasts.