What is perch?
Perch is an open-source semantic linter that checks your code for defects, security issues, and project-specific rules. Instead of relying on pattern matching, it uses SystemOne decision models to understand the meaning of your code. You can define rules like "Store timestamps in UTC" or "Never build SQL from strings" in plain English, and Perch will enforce them alongside its built-in scans.
To get started, you install Perch via npm and run `perch scan` on your project. It checks every method for common problems like off-by-one loops, unchecked empty values, weak hashing, and missing authorization. You can also write your own rules at the project, file, or method level, and Perch will search the most likely places first and stop when it finds an answer.
Perch integrates with Claude Code, Codex, and Cursor, letting your AI agent check its own fixes. For teams, Perch Cloud scans each pull request in its own container, comments on flagged lines, and caches results so unchanged code gets the same answer every time. Pricing starts with a free $5 usage credit, then about 10¢ per 1,000 rule checks, with cached checks free.
perch features
- Built-in scans: Detects off-by-one loops, unchecked empty values, weak hashing, and missing authorization in every method.
- Plain-English rules: Write custom rules like "Store money in integer cents" and enforce them across your codebase.
- Agent skill integration: Installs into Claude Code, Codex, or Cursor so your AI agent can apply rules to its own edits.
- Pull request scanning: Perch Cloud reviews each PR in a container, comments on issues, and caches results for consistency.
- Project, file, and method levels: Run rules at different scopes, from whole codebase questions to per-method checks.
What you can do with perch
- Scan a codebase for security issues like hardcoded keys or weak hashing.
- Enforce project-specific conventions such as storing timestamps in UTC.
- Check that an AI agent's code fixes follow your team's rules.
- Review every pull request automatically without writing CI workflows.
How to get started with perch
- Install Perch globally with `npm install -g @lakeday/perch`.
- Run `perch scan` on your project to check for defects and security issues.
- Sign up, turn on scans for a repository in the dashboard to get pull request reviews.
Tips for better results with perch
- Install Perch globally with npm and run perch scan on your project to catch built-in issues like off-by-one loops and weak hashing before writing custom rules.
- Write custom rules in plain English at the project, file, or method level, and commit them with your code so Perch enforces them consistently across your codebase.
- Use Perch's agent skill to have Claude Code, Codex, or Cursor check its own fixes, turning corrections into rules that apply to later edits.
- Enable Perch Cloud on pull requests to scan each PR in a container and comment on flagged lines, caching results so unchanged code gets the same answer every run.
perch pricing
perch has a free way to start, with paid plans for heavier use. Check perchscan.com for current limits and prices.
What to check before you rely on perch
- Check your free $5 usage credit and per-rule-check pricing on the Perch website, since the exact billing details are not fully listed there.
- Verify that Perch supports your project's language and file types, as the website does not specify all supported platforms or ecosystems.
- Review Perch's data handling and privacy for cloud scans, since the website does not state where your code is processed or stored.
Who perch is for
perch suits developers, engineering teams and AI coding agent users. If that is not you, the AI coding tools below may fit better.
Similar tools compared with perch
| Tool | What it is | Pricing |
|---|---|---|
| perch | Semantic code linter with plain-English rules | Freemium |
| Tabnine | Agentic quality engineering platform for AI code | See website |
| CodeRabbit | AI pull request reviewer and prioritizer | Free trial |
| Cursor | AI coding agent for ambitious software | See website |
| Claude Code | AI coding agent for terminal and IDE | Paid |
perch FAQ
What does Perch do?
Perch is a semantic code linter. It scans your code for bugs, security issues, and project rules written in plain English. It checks every method for common problems and can enforce custom rules you define.
How much does Perch cost?
Perch offers a free tier with your first $5 of usage. After that, it costs about 10¢ per 1,000 rule checks, and cached checks on unchanged code are free. Pull request scans cost about 50¢ per PR.
Who is Perch for?
Perch is for developers and engineering teams who want to catch bugs and security issues early. It's especially useful if you use AI coding agents like Claude Code, Codex, or Cursor, since it can make your agent check its own fixes.
Can I write my own rules?
Yes. You can write rules in plain English, like 'Never build SQL from strings' or 'Refunds can only process once'. Rules can be applied at the project, file, or method level.
Does Perch integrate with GitHub Actions or other CI tools?
Perch Cloud scans pull requests directly from your repository dashboard, so you don't need to write CI workflows. The website mentions GitHub as a supported platform but does not specify integration with GitHub Actions or other CI tools. Check the docs for the latest integration details.
What programming languages does Perch support?
Perch is installed via npm and scans your project's code. The website shows examples in Python and TypeScript, but it does not list all supported languages. To see if your language is covered, check the documentation or examples on the Perch website.
Can I use Perch for free without a credit card?
Perch offers a free $5 usage credit to start, which is enough for about 50,000 rule checks. After that, you pay about 10¢ per 1,000 checks, with cached checks free. The website does not mention whether a credit card is required for the free credit, so check the signup page for details.
