Review every pull request with a panel of specialist agents

SHIP reviews each pull request with separate reviewer agents, such as one for security and one for your conventions. Blocking findings go to a Builder that pushes the fix.

For: Teams that already pay for AI code review.
Instead of: Review comments your engineers still fix by hand.

Runs today: Open a pull request, hand it to SHIP with ship delegate pr, and give that mission's Reviewer extra instructions, such as a security pass, with --prompt-file.

Join the early-access list Read the CLI guide

Answer 4 questions. SHIP emails you when it opens to you.

A Reviewer verdict on a GitHub pull request, signed with the SHIP agent card naming the harness, model and run — A Reviewer verdict on a pull request SHIP built. Review lanes on every pull request are in early access.

How it works

Open a pull request, get one verdict

Step 1

Install SHIP on your repositories

Install the GitHub App and choose the lanes that review every pull request: correctness, security, conventions, or a lane you write yourself.

Step 2

Read findings from every lane

Each lane reviews the diff in its own sandbox on its own model. SHIP merges their findings into one verdict, blocking issues first.

Step 3

Let the agents fix what blocks

A blocking finding goes to a Builder that pushes the fix to the branch. Your CI and the Tester run again before the pull request is ready.

Capabilities

Choose who reviews what

Run the reviewer on a different model from the author

Set each lane's harness and model separately from the Builder's, so a different provider checks the code another one wrote.

Write a lane in your own words

Write a lane as a prompt file in your repository and choose whether it extends your reviewer conventions or replaces them, for one pull request or for all of them.

Test the change as well as reading it

The Tester runs the change on a preview and checks its acceptance criteria, so a review comes with evidence that the change works.

Keep your models and your keys

Every lane runs on the credentials you connect, and your provider bills inference at your own rates.

See what each review cost

Every review reports its cost per lane, priced on the model that actually served it.

Comparison

Compare SHIP with CodeRabbit and Greptile

What each product's own pages say. Last reviewed October 1, 2026; check the vendor's page before you buy.
CodeRabbit	Bills per developer seat, $24, $48 or $72 a month on annual billing, for developers who create pull requests. Pricing Reviews pull requests with multiple AI models. Pull request review Delivers a commit or a follow-up pull request from Autofix and Finishing Touches when someone triggers them. Finishing Touches Connects to your own model provider when you self-host it on Enterprise, for 500 seats or more. Self-hosted
Greptile	Bills per seat; Pro is $30 per seat a month and includes 50 credits per seat. Pricing Builds a graph index of your codebase and runs a swarm of agents on each pull request. Homepage Sends a suggested fix to your own coding agent, which applies it for you to review. Fix with your Agent
SHIP	In early access Runs each review lane on the harness and model you choose, on your own keys. Tests the change on a preview, and sends blocking findings to a Builder that pushes the fix and re-runs your CI. Available now Bills per mission: one credit covers one mission of up to 3 agent-hours, and your own provider bills model usage. Key facts

FAQ

Frequently asked questions

Is multi-agent code review available today?

Partly. Review lanes on every pull request are in early access. Today, run ship delegate pr on an open pull request and SHIP reviews it, tests it and hands it back ready to merge, with --prompt-file adding extra instructions, such as a security pass, to that mission's Reviewer.

Run a lane on one pull request

Does it review pull requests people wrote?

Yes, SHIP reviews pull requests written by people and by agents against the same conventions. Today you hand one over with ship delegate pr after you open it.

When is CodeRabbit or Greptile the better fit?

CodeRabbit or Greptile fits a team that wants AI review on every pull request today, with fixes delivered when someone triggers them or applied by its own coding agent, as their own pages describe. SHIP fits when a blocking finding should come back fixed on the branch, through your CI and a test on a preview, in the same mission.

Get a review panel on your next pull request

Join early access and choose the lanes that review your repositories.

Join the early-access list

Answer 4 questions. SHIP emails you when it opens to you.