NewRepo-aware AI review, now with your own API key

Review instantly

From raw request to merged PR — AI reviews the code, humans approve the release.

Open source · Bring your own key

pipeline · live
feat/csv-export
Request
PRD
Tasks
Build
AI Review
Shipped
An alternative toCodeRabbitQodo MergeGreptileKorbit
Features

Everything between idea and merge

Most tools review a PR after it exists. ShipFlow owns the whole path — and reviews it smarter when it gets there.

Repo-aware reviews

Every review reasons over the diff and the surrounding codebase — not just the lines that changed — so it catches the bugs other tools miss.

Request → merged PR

A raw feature request becomes a PRD, a task breakdown, code, and a reviewed pull request. The whole pipeline, in one place.

Bring your own key

Model-agnostic via OpenRouter. Plug in your own API key, choose your models, and keep full control of spend and data.

Human-approval gates

AI does the review, humans do the deciding. Nothing ships until a real person signs off at the approval gate.

Durable background runs

Workflows are powered by durable execution — a flaky model call retries the step, never the whole pipeline.

Living PRDs & tasks

Editable, AI-drafted product specs and task boards stay in sync with the work as it moves toward production.

Core loop

Request to production — one pipeline

Five phases. One observable loop. Click a phase or watch it cycle.

core loop · live

Phase 1

Product discovery

Intake requests, clarify with AI, educate if the feature already exists.

Request → PRD → Tasks → Code → Review → Ship
How it works

One loop, fully observable

From the first message to the merge button — here is the journey every feature request takes.

  1. 01

    Drop in the request

    Paste a customer email, a ticket, or a one-liner. ShipFlow asks clarifying questions when it needs more context.

  2. 02

    AI drafts the plan

    A structured PRD and a task breakdown are generated automatically — fully editable before anything is built.

  3. 03

    Review with full context

    When the PR opens, ShipFlow reviews it against the whole repo, flags real issues, and suggests concrete fixes.

  4. 04

    Approve & ship

    A human signs off at the approval gate. ShipFlow routes it to merge and marks the request shipped.

#128 · feat: csv export
AI reviewed

+ export function toCsv(rows: Row[]) {

- return rows.map(r => r.join(","))

ShipFlow review

Values containing commas or quotes will break the output. Escape fields and wrap in quotes before joining.

Fix accepted· Ready to merge
Why ShipFlow

More than a PR commenter

CapabilityShipFlowTypical tools
Repo-aware review contextpartial
Request → PRD → tasks pipeline
Bring your own model keypartial
Human-approval gates
Durable background workflows
Pricing

Simple, honest pricing

Start free. Bring your own key on any plan. Upgrade when your team ships more.

Free

For solo devs kicking the tires.

₹0/ month
Get started
  • 1 connected repository
  • 5 AI reviews / day
  • 25 AI reviews / month
  • Feature request → PRD → tasks
  • Bring your own API key
Most popular

Pro

For teams shipping every day.

₹599/ month
Start Pro
  • Unlimited repositories
  • Unlimited AI reviews
  • Repo-aware review context
  • Priority pipeline runs
  • Human-approval gates

Business

For orgs with scale & compliance needs.

₹1,599/ month
Start Business
  • Everything in Pro
  • SSO / SAML
  • Self-hosted model routing
  • Audit logs & SLAs
  • Dedicated support
FAQ

Questions, answered

Both. ShipFlow runs reviews directly on your pull requests via the GitHub App, and gives you a web dashboard to manage feature requests, PRDs, tasks, and the whole shipping pipeline in one place.

Those tools focus on reviewing a PR once it exists. ShipFlow reviews with full-repository context AND drives the work that comes before the PR — turning a raw request into a PRD, breaking it into tasks, then reviewing the resulting code and routing it for human approval.

Yes. ShipFlow is model-agnostic via OpenRouter. Drop in your own key and pick the models you want for drafting vs. reviewing. Your key, your spend, your data boundary. See the Bring-your-own-key docs.

No. Your repositories are used only to generate reviews and artifacts for you. Nothing is used for model training. With your own API key, requests go directly to the provider you choose.

Any repository you can connect through the GitHub App — public or private. The review reasons over diffs and surrounding repo context, so it works across stacks.

Under ten minutes: sign in, connect the GitHub App to a repo, link it to a project, and open your first feature request. The quick-setup guide walks through every step.