AI review

AI code review on your own provider — including local models.

Configure a supported CLI, a localhost model, or an OpenAI-compatible endpoint. PR Flow sends each AI task to the provider you select; a localhost endpoint keeps the prompt on your machine.

Download freeRead the docs
AI insights
PR Flow AI insights panel showing a review summary, complexity and risk estimate and inline findings
Start with the data path

Know where each AI request goes.

Each provider row names its destination before you run a review. Choose a local model, an authenticated CLI, or your own compatible endpoint.

No hosted PR Flow model

PR Flow composes the request on your machine and sends it to the provider you configured.

Someone else's model choice

Select the provider and model for each task, including an approved internal gateway.

Score theater

Findings point to code and estimates remain advisory. The human reviewer decides what is useful.

Bots that act on their own

Findings and discussion answers can become staged comments. They never post autonomously, and they do not approve or merge.


Bring your own provider

Supported providers, routed per task.

Configure providers under Settings → Integrations → AI providers. Each row names where prompts go and includes a connection test.

Built-in CLIs

The Claude, Antigravity, and Codex command-line tools use their existing authentication.

A local model

Ollama or another compatible service on localhost keeps prompts on your machine.

An OpenAI-compatible endpoint

Supply the URL and key for a compatible service or internal gateway. The key is stored encrypted.

A custom local CLI

Configure a command and prompt argument for a compatible command-line tool.

Task routing: mix them

Route deep review, estimates, thread summaries, Discuss, and summary drafts separately. Results identify the provider and model that produced them.

What you get back

Findings on lines, not a grade out of ten.

In the review view

Pins you accept or dismiss.

Findings whose quoted code still matches the file appear as gutter pins. Accept one to create a staged comment or dismiss it. The comment reaches the code host only when you submit the review.

  • Advisory summaries, complexity, risk, and effort from enabled helpers
  • Estimates are marked outdated when the pull request changes
1 finding Quick accepted
Review · acme/web-app #101
PR Flow review view with an AI finding pinned to a diff line and a staged comment
Change Story

Read the change in the order it makes sense.

File order puts the caller before the contract and the test three files from the behaviour it proves. Ask for a story and the same diff is ordered into stops — one idea at a time, gathering the hunks that belong together even when they sit in different files. Each stop says what changed and what to check; it never says the code is correct.

  • Everything left off the path is listed, with the reason, and the counts match the file view
  • The ordinary diff stays one click away, and your comments and drafts are untouched
  • Stops you have opened are counted separately from files you marked reviewed
Story · acme/web-app #101
PR Flow Change Story showing a stop above the diff, its walk-through expanded, and the chapters and stops listed in the rail
Discuss

Ask about a file, a finding, or the whole change.

Ask your configured provider about a file, finding, or the pull request. An answer can become a staged comment; it still needs your review and submission.

  • Scope the conversation to a file, a finding, or the whole pull request
  • Runs on a CLI, a local model, or your gateway — same routing rules
  • Turn any answer into a comment you still have to submit
Discuss · your AI
PR Flow discuss panel answering a question about a diff using your configured AI provider
Agent Pipelines

Every prompt has an explicit context allowlist.

Each agent pairs a prompt with the pull-request context it may receive. You can inspect and change that selection before it runs.

What an agent may see

You choose, per agent.

  • Diff — the pull request's code diff
  • PR details — title, description, changed files, open threads
  • Linked ticket — a referenced Jira, Linear or Trello item
  • Git history — recent commits on the changed files
  • A named file — a specific path at the PR's latest commit

Agents produce findings or notes for the reviewer. They do not change code, commit, push, approve, merge, or resolve conversations.

Community agents show their full prompt and exact context before anything is saved, and an approved version keeps running until you review an update. The Security Reviewer, Performance Reviewer and Test Reviewer are good examples of how narrow a useful agent can be — browse the rest in the Agent Gallery.

PR detail · Agents
PR Flow AI insights card showing a custom agent's inline findings on a pull request
Verify it yourself

Logs you own, on your machine.

Check AI runs and write actions from Settings → Logs on your machine.

  • Settings → Logs → AI runs — metadata for every run, exportable as JSON. Storing the prompts themselves is an explicit opt-in.
  • Settings → Logs → Write actions — every write PR Flow made for you, with a link to the result

The full picture is in the privacy policy and security overview, and the setup detail is in the AI features documentation.

Settings · Logs
PR Flow's local AI run log, listing each run with the provider and model that produced it
Questions

Where the code goes.

Where does my diff actually go?
Wherever you point it. PR Flow does not host a model and does not proxy prompts — it composes the request on your machine and sends it to the provider you configured. If that is Ollama on localhost, the prompt never leaves your machine. If it is your company gateway, it goes to your gateway. If it is the Claude CLI, it goes wherever that CLI is authenticated to send it. Every provider row in Settings shows a plain-English data-flow label saying which of those applies.
Can I run code review entirely offline?
The AI half, yes — point PR Flow at a local model server such as Ollama, LM Studio, llama.cpp, vLLM or LocalAI and prompts to that localhost endpoint stay on your machine. PR Flow still needs network access to read pull requests from GitHub, GitLab, Gerrit or Azure DevOps, since that is where they live.
What happens if I never turn AI on?
Nothing is sent anywhere. If you never enable an AI feature, none of your pull request content reaches any AI provider. The queue, the ranking, the review view and submit all work without it — AI adds summaries, estimates, inline findings and the guided reading path on top.
Can the AI approve or merge something?
No, and this is structural rather than a setting. Everything an agent produces is advisory — findings pinned to lines, a freeform note, or a reading path through the diff — and nothing else. It cannot post, approve, request changes, merge, run commands or browse your machine. A finding only becomes a comment when you accept it, and that comment only posts when you press submit. The one thing that reaches beyond the pull request is repository context for the reading path, and only when you have mapped that repository to a local checkout and separately granted permission: PR Flow reads those files itself, at a fixed revision, and passes the result along as text.
Can a team force everyone onto one provider?
Yes. A team license can lock AI to specific providers, enforced on every seat — so “code only goes to our own gateway” is a policy rather than a setting each person has to remember. The team admin manages this from the license portal; see pricing for how Team plans work.
Where are API keys stored?
For hosted endpoints you add, the API key is stored encrypted and is never written into your settings file. Credentials you enter anywhere in PR Flow are encrypted at rest using your operating system's secure storage — Keychain on macOS and the equivalent on Windows and Linux — and leave your machine only to authenticate with the service they belong to.
v1.18 · macOS · Windows · Linux

Keep AI advisory and under your control.

You still review the code and decide what to submit. Try PR Flow free for 14 days; a one-time licence keeps it after the trial.

Download freeSee pricing