Source: https://codereviewguide.fyi/open-source-code-review-cli/

# Open-source AI code review CLIs: what can you run and change?

Perch has an MIT-licensed CLI, saved English requirements, and checks that read uncommitted files. PR-Agent is a separate MIT project with local Git and pull-request review. Choosing open-source software also means choosing where model inference runs and who pays for it.

| Question | Perch | PR-Agent |
| --- | --- | --- |
| What is open source? | The Perch CLI is MIT licensed. | The community PR-Agent project is MIT licensed; it is separate from managed Qodo. |
| What does local review read? | `perch check` reads current working files; a file::method target selects a method. | The local Git provider reviews committed branch changes against a target branch and requires a clean working tree. |
| How are requirements configured? | Versioned YAML with named project, file and method requirements. Select a named rule when rerunning a check. | Repository configuration and review instructions, with settings for the selected review tool and model. |
| Does the license include model inference? | Hosted model analysis is a paid service; relevant source is sent to the configured model service. | Configure the model provider and account. Hosting and inference depend on that deployment. |

## Run the saved requirement before committing

Use a requirement tied to the failure your agent keeps introducing. The FastAPI tenant example keeps the authenticated tenant inside the same atomic update that changes the order. Its rule pack contains the requirement and agent task; the tutorial links the broken and fixed source checks.

[FastAPI tenant authorization: implementation and captured checks](https://frameworkrecipes.dev/fastapi/tenant-cancellation/) · [Download the tenant authorization rule pack](https://reporules.fyi/fastapi/tenant-cancellation/)

Install and configure the agent you use:

```
npm install -g @lakeday/perch
perch login
perch setup claude-code
# Or: perch setup codex
# Or: perch setup cursor
```

Download the pack without replacing existing files. Its example selector targets orders.py; adapt that selector to your application before checking.

```
curl -fL 'https://reporules.fyi/fastapi/tenant-cancellation/rule-pack.zip' -o perch-tenant-rule.zip
unzip -n perch-tenant-rule.zip
perch check orders.py --rules tenant-scoped-update --json
# Repair the finding, then repeat the same check:
perch check orders.py --rules tenant-scoped-update --json
```

After installing a focused pack, inspect its selector and use its exact named check command after the edit. Read the finding, repair the implementation, and repeat the same command. Keep your existing perch.yaml; the pack instructions use extraction that preserves existing files. Fresh judgments can vary; unchanged requests reuse cached answers.

## Separate developer seats from model usage

Perch's hosted rate is 10¢ per 1,000 fresh rule checks, with cached checks free and no developer seat fees. PR-Agent's open-source license does not determine your model-provider bill. Compare the configured scope, inference charges, infrastructure and review frequency.

[Worked review costs for three and ten developers](https://reviewpricing.fyi/team-review-costs/)

## Check commercial CLI alternatives against the same requirements

CodeRabbit documents local committed and uncommitted review. Greptile documents committed-branch CLI review. Their available CLIs do not establish an open-source license for the review service. The dedicated comparisons cover current plan allowances and custom-rule workflows.

[Perch vs CodeRabbit: CLI, saved rules and billing](https://codereviewguide.fyi/perch-vs-coderabbit/) · [Perch vs Greptile: working files and committed branches](https://codereviewguide.fyi/perch-vs-greptile/)

## Inspect the evidence and source documentation

The public evaluation includes Perch's recorded results on four disclosed requirements. CodeRabbit and Greptile remain unscored until their authenticated reviews complete. This licensing comparison does not establish comparative detection accuracy.

[Matched public inputs and recorded review results](https://reviewbenchmarks.dev/evaluations/order-service/)

- [Perch MIT license](https://github.com/lakeday-org/perch/blob/main/LICENSE)
- [Perch working-file and method checks](https://github.com/lakeday-org/perch/blob/main/docs/check.md)
- [Perch named requirements and selectors](https://github.com/lakeday-org/perch/blob/main/docs/rules.md)
- [PR-Agent project and MIT license](https://github.com/The-PR-Agent/pr-agent)
- [PR-Agent local Git provider](https://docs.pr-agent.ai/usage-guide/local_git_provider/)
- [CodeRabbit CLI documentation](https://docs.coderabbit.ai/cli)
- [Greptile CLI documentation](https://www.greptile.com/docs/code-review/greptile-cli)

Reviewed October 14, 2026. Published by Perch / Verglas LLC.

**Code Review Guide**

Tools & workflows.

Published by Perch · © 2026 Verglas LLC. Code Review Guide is a Perch publication.

Perch’s MIT license covers the CLI; hosted model analysis is a paid service. Relevant source is sent to the configured model service.

[Perch CLI MIT license](https://github.com/lakeday-org/perch/blob/main/LICENSE) · [Article text](https://codereviewguide.fyi/open-source-code-review-cli/index.md) · [Resource index](https://codereviewguide.fyi/llms.txt)
