Commands
Ask Angada for an AI review from a GitHub pull request.
Post a top-level comment in the pull request's Conversation tab to ask Angada for a review or help.
Use @angada-ai review to start a review. @angada-ai help only lists commands; it never starts one.
See the command flow from your perspective
flowchart TD
A["Open the pull request Conversation tab"] --> B["Post a top-level @angada-ai comment"]
B --> C{"Are you on the workspace roster or the personal installation owner?"}
C -- No --> D["Angada reacts with 😕; no public reply"]
C -- Yes --> E{"Which command did you post?"}
E -- help or question mark --> F["Angada replies with available commands"]
E -- Unknown or planned command --> G["Angada reacts with 😕 and points you to help"]
E -- review or bare mention --> H{"Can Angada review this pull request head?"}
H -- No --> I["Angada explains what is blocking the review"]
H -- Yes --> J["Angada reacts with 👀 and reviews the current head"]help, unsupported commands, and unknown commands do not start a review, create a check run, or use review credits. Review requests can be held by permissions, review settings, limits, or the pull request's current state; see Troubleshoot a review for next steps.
Commands that work today
| Comment syntax | What Angada does |
|---|---|
@angada-ai review | Starts a full review of the current pull request head when it is eligible. |
@angada-ai | Starts the same full review as review. |
@angada-ai help | Lists the available commands without starting a review. |
@angada-ai ? | Shows the same command list as help. |
Use the exact handle @angada-ai; @angada belongs to another GitHub account. Angada reads the token immediately after the handle on the same line, so it does not parse a natural-language request. Text after a supported command is ignored. A reply inside an inline review thread does not run a command; post in the pull request's Conversation tab.
Who can use each re-run option
| Where you request the review | Who can request it | What to expect |
|---|---|---|
| Pull request comment | A member of the Angada workspace roster. The owner of a personal GitHub installation is also eligible. | Post @angada-ai review on a new, eligible head. |
| Angada web app or REST API | A workspace admin. | Request a manual review for an eligible pull request. |
| GitHub Checks | A GitHub user allowed to re-run that check by the repository's GitHub permissions. | Choose Re-run on the Angada check. |
Each path requires a current pull request head that has not already been reviewed. Workspace roles control Angada actions; GitHub permissions control the Re-run action in Checks. See Re-runs and continuity for how these paths differ.
What happens after a mention
When Angada accepts an eligible review request, it reacts with 👀 and starts a full review of the current head. If the request cannot start, Angada gives an eligible commenter the applicable reason. It can name a closed pull request, a review already running or completed for that head, the mention limit, the pull request or Workspace review cap, exhausted credits, no Review Seat for the pull request author, a skipped-author rule, AI Review being turned off, or GitHub permissions awaiting approval.
Only a workspace roster member can use mention commands, with the personal installation owner as an exception. If someone outside those rules mentions Angada, it reacts with 😕 and does not post a public refusal comment. A roster member who uses an unsupported or unknown command gets a 😕 reaction and a reply that points to help. Neither help nor an unsupported or unknown command starts a review, creates a check run, or uses review credits.
Limits and drafts
- Angada accepts at most three mention-triggered review dispatches for a pull request in a rolling 24-hour window.
- Mention requests still follow the shared pull request and workspace review caps, credit limits, and Review Seat rules. See Limits and caps and Review settings.
- Angada does not review the same head twice. If a review for that head is already running or complete, push a new commit before requesting another review.
- A manual mention can review a draft pull request even when automatic draft reviews are off. It records a per-PR override, so later heads of that draft are also eligible for automatic review.
Upcoming commands
These command ideas are planned and are not available yet. Each is intended to make it easier to work through a review directly from the pull request:
| Command | What it will do |
|---|---|
why | Explain why a finding was raised and point to the evidence behind it. |
fix | Prepare a suggested fix for a finding so you can review the change. |
ignore | Dismiss a finding that does not apply to your code. |
deep | Ask Angada to take a closer look at the pull request. |
explain | Walk through a finding in more detail, including its impact. |
The command list may grow over time. For commands available today, post @angada-ai help in the pull request conversation.
Responding to findings

Review findings appear alongside the changed code, suggested changes, and discussion on the pull request.
Use the pull request review controls to resolve a finding after fixing it, or dismiss it when it does not apply. Angada carries those outcomes forward when it matches the same finding on a later pull request head. This helps keep follow-up reviews focused; it does not add a command or change how future findings are generated. See Understand a review for how findings and their outcomes appear across pushes.
Troubleshooting commands
| What you see | What to check |
|---|---|
| No reply, or only a 😕 reaction | Confirm you used the exact @angada-ai handle and posted in the pull request's Conversation tab. Mention commands are available to workspace roster members and the personal installation owner. |
A reply points you to help | The command is not available yet or Angada did not recognize it. Post @angada-ai help to see the supported commands. |
| Angada says the head was already reviewed | Push a new commit, then request a review of the new head. |
| A review cannot start | Check the reason in Angada's reply, then see Troubleshoot a review for permissions, settings, limits, and credits. |
To check the GitHub permissions AI Review needs, see GitHub integration.