HyperNeo Review

Your review, your rules

Most AI reviewers ask you to trust their engine, their model, their billing. This one is yours to configure:

• Bring your own key. Your provider API keys, your provider account — when you bring them, inference runs on your key and your billing, under your control.
• Any harness. The review runs on the agent harness you choose — Claude Code, Codex, and more. How the reviewer works is configuration, not a black box.
• Any provider. Claude, Codex, DeepSeek, and more — pick the primary engine per repository, run shadow engines alongside, change your mind whenever.

The trust question is answered in the open: review-bench publishes the benchmark methodology, schemas, and corpus — first batch landed, results follow as runs do — and methodology critique is welcome in the forum.

Pricing

Bring your own keys — pay for the harness, not the tokens. Inference runs on your provider keys, so there is nothing to meter: no per-review charges, no token billing, no surprise invoices.

• Free

  $0

  • Public repositories
  • 14-day no-card trial on private repositories
  Install the App

• Pro

  $15 / developer / month
  $12 billed annually

  • Private repositories
  • Bring your own provider keys — sealed at rest, inference on your billing
  • Any harness per repo: Claude Code, Codex — more harnesses rolling out
  • Any provider: Claude, Codex, DeepSeek, and more
  • Review timestamps in the display timezone you set
  • Self-host friendly — one container image, SQLite storage
  Get Pro

• Design Partner

  $7.50 / developer / month

  • Everything in Pro
  • Lifetime 50% off — locked for early users
  Become a design partner

Card-free soft open: no card is required to start, and design partners are invoiced manually. When checkout goes live, these buttons wire straight to payment links — until then, email hello@hyperneo.dev and design-partner pricing is locked for early users.

Quickstart — 15 minutes to first review

1. Install the GitHub App. github.com/apps/hyperneo-ai — grant access to the repositories you want reviewed. Repos you grant are enrolled automatically and start advisory.
2. Sign in. agent.hyperneo.dev/login with GitHub OAuth (the installation account owner or a collaborator on an enrolled repository). Operators land in the admin console; everyone else in the customer app.
3. Add your provider key. In the app under Keys: pick your provider, paste the API key, save. Keys are sealed — stored encrypted at rest and shown afterwards only as a label and fingerprint — and your reviews run on your provider account.
4. Open a pull request. The review posts each finding as an inline comment at its diff position with severity and confidence, and finishes with a one-line verdict naming the engine and model that reviewed.

No card at any step: public repositories are free and the private-repo trial runs 14 days without one.

If you are an AI agent maintaining a repository

What this does

On every pull request to an enrolled repo: reviews the diff, posts each finding as an inline review comment at its diff position — severity (P0/P1/P2; P0 and P1 are blocking, P2 is a minor note) and confidence — and finishes with a one-line verdict: “Approved.” or “Requires changes.” The review body names the engine and model that reviewed, the commit reviewed, and the trigger. A status-table comment tracks each run: started/finished, head, engine, verdict. Reviews are advisory by default; humans merge.

Setup — human, once, ~15 minutes

1. Install the GitHub App: github.com/apps/hyperneo-ai (Marketplace listing pending)
2. Grant access to the repositories to review
3. Sign in with GitHub OAuth (the installation account owner or a collaborator on any enrolled repository): agent.hyperneo.dev/login — operators land in the admin console; everyone else in the customer app (/app)

Enroll a repository

Admin UI → Repos → Enroll → owner/name plus the installation id (the GitHub App installation; required with App auth). Or simply grant the App access to the repo in your GitHub settings — auto-enroll follows, no id needed. New repos start advisory, primary engine claude, triggers open/push/mention. Reviews fire on the next PR open/push.

Trigger a review

• Automatic: PR opened, reopened, or marked ready for review; every new push. Draft PRs are skipped until ready. Newest head wins — an in-flight review folds into the new head.
• On demand: comment on the PR mentioning @hyperneo-ai (e.g. @hyperneo-ai review) — needs the mention trigger on (it is by default). Any extra text in the comment steers that review.

All triggers honor the repo's trigger settings; open, push, and mention are on by default.

The review contract

• Findings: inline comments at the exact diff line — severity, confidence, what and why
• Verdict: one line in the review body, with engine/model provenance; clean PRs get the verdict and provenance lines only
• Findings on lines outside the diff: bare file:line — severity pointers in the body
• 👀 while running; removed when the run finishes

Configure per repo (admin UI)

Mode — advisory (default; comments only) or gate (approve / request changes) · primary engine + shadow engines (shadows run alongside, are recorded, and are not posted) · triggers (open / push / mention / manual) · diff excludes · extra guidance (steers what the reviewer looks at)

Agent etiquette

• Address findings by pushing fixes; the next push is re-reviewed automatically when the push trigger is on (it is by default)
• The verdict is advisory to the human who merges
• The bot does not converse — don’t reply to its comments

Machine access

This manual: https://hyperneo.dev/agent.md (text/markdown, stable URL)