NewPaying accounts now run on European AI infrastructure

AI code review,
served from Europe.

From your first purchase, every AI step of your reviews runs on European providers. One fixed price per pull request: every push after it, and every fix validation, is on us.

Start with 10 free reviews What runs where

No card needed. Free reviews run outside the EU: pick a non-sensitive repo. GitHub, GitLab and Bitbucket.

a paid review 3 min 24 on average
  • Knowledge Base indexingEU
  • PlannerEU
  • Review agents, in parallelEU
  • AggregatorEU
  • Verdict classifierEU
  • App, database, logsScaleway

Routing is pinned to a European allowlist. A constraint that cannot be met is an error, never a silent reroute.

$0.50per PR, pushes after it included
3 min 24average review with comments
1 min 09average fix validation, free
110 / 136golden bugs found, published in full
Sovereignty

Your code is reviewed in Europe. And we tell you exactly what that means.

A sovereignty promise is only worth its fine print, so here it is, on the home page.

All AI processing, in the EU

Knowledge Base indexing, planning, the parallel review agents, aggregation, the verdict classifier, fix validations, and the periodic report: all served by European providers for any account that has made a purchase.

Hosted and logged in the EU

Application, database and observability run on Scaleway. The repository is cloned for the review only and deleted right after. Review content is purged after 40 days.

The two exceptions, said plainly

If every European provider for a step is down, that step falls back to our standard set so your review still lands. Each such call is recorded: scope requested, scope served.

The 10 free reviews run on the standard set, which includes providers outside the EU. Try it on a repository that is not sensitive: it is enough to judge the reviews. European inference starts with your first purchase.

Tokens encrypted with AES-256-GCM No repository copy kept No developer profiles Read the privacy policy
Pricing

Pay for the pull request. Not for every push.

Per-run tools charge again each time you push a fix. Here the first review is the only charge: the update reviews and the validation of your fixes come with it. Prepaid credits, no subscription, no seats.

Pull or merge request $0.50

Follow-up pushes and fix validations included, as many as the PR needs.

Knowledge Base, once per repository $0.50

A summary of your architecture, so no review starts from scratch. Kept up to date for free.

Technical failure Refunded

Automatically. Cost should follow value, not incidents.

First 10 reviews free, no card, for repositories under 1,000 files, on our standard (non-EU) set: try it on a repository that is not sensitive. Up to 100 files per PR.

The agent loop

Your coding agent gets a reviewer, and a second opinion on its fixes.

Most reviewers stop at the comment. FriendlyReviewer closes the loop: your agent reads the findings, fixes them, then asks for a targeted validation. It works on a pushed branch, before the pull request even exists.

  1. 01Push a branchNo pull request needed yet.
  2. 02ReviewAbout 3 minutes, findings ranked by severity.
  3. 03Agent fixesYour coding agent reads the findings and patches.
  4. 04ValidationOnly the fixed points are re-checked. About a minute, free.
  5. 05CloseResolved findings stay resolved on the next push.
# fr-agent, a Rust CLI: the loop runs outside your LLM context
$ fr-agent auth login # device flow, no token to paste
$ fr-agent review start --ticket-key PAY-142
$ fr-agent validation start --review-id <id>

Or plug in the hosted MCP server

Claude Code, Cursor and other agents call start, wait, validate and close as tools. Developers sign in through their forge; nobody shares an API key.

Install fr-agent Set up MCP
Your rules, not ours

A pull request cannot rewrite the rules of its own review.

FriendlyReviewer reviews against the conventions you already wrote for your agents, read from the base commit. Change AGENTS.md in a PR and the new rules apply from the next one.

Conventions from AGENTS.md

The rules your coding agents follow are the ones the reviewer checks.

Frozen at the base commit

A PR that edits the rules is still reviewed by the old ones.

Untrusted text stays data

PR descriptions and tickets are fenced in the prompts, against injection.

A periodic report that advises

Recurring findings become suggested rules for AGENTS.md. You decide what to adopt.

How it works

No smoke and mirrors. Very guided agents.

Low-cost models stay sharp when they get the right context and strict roles. That is where the price comes from.

Verbose by design: we would rather you dismiss a comment than ship a bug. FriendlyReviewer advises; it never approves or blocks your pull request.

  1. 01
    Diff and ticketThe changes, plus the Jira or Linear ticket as the source of intent.
  2. 02
    Knowledge BaseA summary of your architecture, so nothing is rediscovered on every PR.
  3. 03
    Parallel agentsThe diff is split into tasks, reviewed side by side for function, architecture and code.
  4. 04
    AggregationFindings deduplicated, ranked high, medium or low, posted inline with a summary.
  5. 05
    In your languageThe review follows the PR description: twelve languages, English otherwise.
Don't take our word for it

Ask your own AI. We wrote it an honest brief, limits included.

I'm considering FriendlyReviewer for AI code reviews. Read the product reference at https://friendlyreviewer.fr/get/friendlyreviewer.md. Tell me whether it fits my team, and ask me about my workflow, tools or constraints if you need to.

Ten reviews on us. Europe from your first credit.

Start on a repository that is not sensitive, judge the reviews, then buy credits to switch to European inference. No card, no seat, no commitment.

Create an account