CodeHawk FAQ
What is CodeHawk?
CodeHawk is an AI-powered GitHub App that automatically reviews pull requests. When a PR is opened, CodeHawk analyzes the code diff using Claude and posts inline comments flagging bugs, security issues, missing error handling, and other problems.
It's designed to catch the mechanical layer of code review — the obvious bugs that humans miss when they're tired or moving fast — so your human reviewers can focus on architecture and design.
How does CodeHawk work?
- You install CodeHawk from GitHub Marketplace
- A developer opens a pull request
- GitHub triggers CodeHawk's webhook
- CodeHawk fetches the diff and sends it to Claude for analysis
- Claude identifies issues and suggests fixes
- CodeHawk posts inline comments on the PR
- Developer sees specific feedback on the exact lines with problems
No config needed. CodeHawk works out of the box.
What does CodeHawk catch?
CodeHawk flags:
- Security issues — SQL injection, XSS, missing auth checks, exposed secrets
- Logic errors — null pointer dereferences, off-by-one errors, race conditions
- Missing error handling — unhandled promises, uncaught exceptions
- Performance problems — N+1 queries, inefficient loops
- Style inconsistencies — deviations from your team's conventions
It works on all languages: TypeScript, Python, Go, Java, Ruby, JavaScript, and more.
What doesn't CodeHawk catch?
- Architecture decisions — whether this feature should be a microservice
- Business logic — whether the algorithm implements the spec correctly
- Codebase conventions — unwritten rules specific to your team
- Intent — whether the code is maintainable or clear to the next reader
These require human judgment. CodeHawk is not a replacement for human review — it's a complement.
How fast is CodeHawk?
Most reviews post within 5–30 seconds of PR open, depending on diff size.
Very large diffs (>10,000 lines) may take up to 2 minutes. You can configure CodeHawk to skip certain paths to speed things up.
Does CodeHawk store my code?
No. Here's the flow:
- CodeHawk fetches the diff from GitHub
- CodeHawk sends the diff to Claude's API for analysis
- Claude analyzes it and returns issues
- CodeHawk posts the issues to GitHub
- The diff is not stored — it's deleted after analysis
Your code never persists in CodeHawk's servers. If you're self-hosting, you control where the data goes.
What's the pricing?
During beta: Completely free, no limits, no credit card required.
When we exit beta:
| Plan | Price | Reviews/month | Per repo | Use case |
|---|---|---|---|---|
| Free | $0 | 3 | Any | Try it out, small teams |
| Pro | $79/month | Unlimited | Any | Active development teams |
Beta pricing advantage: You get unlimited reviews right now. Beta users will get priority for discounted pricing after launch.
Pro tier (post-beta): $79/month per organization (not per repo). Covers unlimited reviews across all your repos.
Billing will be managed through GitHub Marketplace.
Can I customize CodeHawk's behavior?
Yes. Create a .codehawk.yml file in your repo root:
# Only flag errors, skip warnings and info
severity_threshold: error
# Ignore generated code and dependencies
ignore_paths:
- "*.generated.ts"
- "dist/**"
- "node_modules/**"
- "vendor/**"
# Focus on specific issue types
focus_areas:
- security
- error-handling
All fields are optional. The defaults work for most teams.
Can I dismiss or ignore CodeHawk comments?
Yes. Every comment has a Resolve button. You can dismiss CodeHawk's feedback if you disagree.
You can configure CodeHawk via .codehawk.yml to adjust severity thresholds and focus areas, reducing false positives over time.
What if CodeHawk is wrong?
CodeHawk is wrong sometimes. It might flag something that's intentional, or miss something subtle.
Dismiss the comment. You can tune CodeHawk's behavior via .codehawk.yml to reduce noise.
You can also configure CodeHawk to focus on specific issue types (via .codehawk.yml) to reduce false positives.
Can I turn CodeHawk off for specific repos?
Yes. When you install CodeHawk, you choose which repos it applies to. You can:
- Install on all repos
- Install on specific repos only
- Uninstall entirely and reinstall later
You can also change repo settings in GitHub's app settings page.
Does CodeHawk work with my CI/CD pipeline?
CodeHawk posts inline PR comments, which integrates with GitHub's normal PR review flow.
You can configure CodeHawk to request changes on error-level issues (which blocks merging) or comment on warnings (which don't block).
Set this via .codehawk.yml:
# Request changes on errors, comment on warnings
error_action: request_changes
warning_action: comment
Is CodeHawk open source?
CodeHawk is closed source, but you can self-host it.
We provide a self-hosting guide with Docker and Node.js instructions. You need:
- Node.js 22+
- An Anthropic API key
- A GitHub App registration
What's the difference between CodeHawk and human code review?
| Aspect | AI (CodeHawk) | Human Reviewer |
|---|---|---|
| Speed | Seconds | Minutes to hours |
| Mechanical bugs | Excellent | Good |
| Architecture | Poor | Excellent |
| Business logic | Poor | Excellent |
| Consistency | Perfect | Varies |
| Fatigue | Never | Always |
| Cost | $79/mo | Salary / time |
Best practice: Use both. CodeHawk handles the mechanical layer. Humans handle architecture and design.
How do I get support?
Email: support@codehawk.crossgen-ai.com
Or open an issue on GitHub Discussions.
Can I use CodeHawk in a private GitHub Enterprise instance?
Yes, but you'll need to self-host CodeHawk. See the self-hosting guide.
What languages does CodeHawk support?
All of them. CodeHawk works on any language — TypeScript, Python, Go, Rust, Java, C++, JavaScript, Ruby, PHP, C#, and more.
Is there a Slack integration?
Not yet, but you can:
- Use GitHub's native Slack integration to post PR notifications
- CodeHawk's comments will appear in those notifications
A native CodeHawk → Slack integration is on the roadmap.
Can I integrate CodeHawk with my linter?
Not directly, but CodeHawk complements linters:
- Linters check syntax, style, and simple patterns
- CodeHawk checks logic, security, and error handling
Use both. They catch different things.
What data does CodeHawk collect?
CodeHawk collects:
- How many reviews per org (for billing)
- Issue counts and types (for product analytics)
- Webhook timestamps (for monitoring)
CodeHawk does not collect:
- Your code (deleted after analysis)
- Personal data
- Repository contents
See our Privacy Policy for details.