Code review

Code Review on Every Push

Turn on review with one organization-wide switch and every pull request gets a review of its new commits, posted as inline comments. Add a pipeline file to scope which pull requests are reviewed, write your own reviewers with their own prompts and models, and cap what a review can spend.

Comments posted by severity
Last 30 days · your-org · 1,068 comments
Median review 2m
4003002001000
58
196
372
318
124
Blocking
High
Medium
Low
Trivial
01

Only the new commits, anchored to the reviewed commit

Each push gets a review of the commits since the last one, never the whole pull request again, and it never re-comments a line it already covered. Comments anchor to the reviewed commit, so GitHub collapses them as outdated when a later push changes those lines.

api-repo #512never re-comments a reviewed line
push 13 commits
reviewed a4f1…9c022 comments
push 21 commit
reviewed 9c02…e7b81 comment
push 32 commits
reviewed e7b8…1d40no findings

Only the commits since the last review

02

Reviewers with the repository checked out

Each reviewer runs in its own sandbox with the repository at the reviewed commit, so it judges a change against the surrounding code rather than the diff alone. Findings post as one review, highest severity first.

# reviewing api-repo#512, push 2

$ pytest -k trade

14 passed

$ git diff HEAD~2 -- src/trade_engine.py

reads the actual change, runs the actual suite, then comments

Judged against the surrounding code, not the diff alone

03

Write your own reviewers

A pipeline file in the repo declares reviewers with their own prompts, models, and path filters, such as a migration-safety specialist that only runs when a migration changes. Reviewers run in parallel, and a budget caps what one run can spend.

agents/code_review.yamllive on merge
ellipsis:  version: v1  kind: code_review  name: Code review  description: Review the new commits on every pull requestpull_requests:  repositories: [web-repo, api-repo]  for:    users: true    bots: falsereview:  - name: correctness    harness:      type: claude_code      model: claude-haiku-4-5-20251001    instructions: |      Flag correctness bugs, unhandled edge cases, and      breaking changes. Suggest the fix on the exact line.budget:  run: 10.00

Your prompts and your models, committed to the repo

You can use Ellipsis for...

Every use case runs on the same platform. See how it works