AI code reviewers you configure with bullet points

What it does

Continue runs automated code review agents on every pull request as native GitHub status checks. Each agent is a markdown file stored in .continue/checks/ inside your repository; it gets a name and description via YAML frontmatter, followed by plain-English instructions like “No secrets hardcoded.” The agent marks the check green if the code looks good, or red with a suggested diff if it finds problems.

The interesting bit

The entire agent definition is just a markdown file with bullet points—no custom DSL or JSON schema to learn. Because the checks live in the repo, your review rules are version-controlled, diffable, and forkable like any other code. An open-source CLI called cn executes the agents.

Key highlights

• Review rules are plain markdown with YAML frontmatter, stored under .continue/checks/
• Integrates with GitHub as first-class status checks that post pass/fail results and suggested diffs
• Cross-platform open-source CLI (cn) distributed via install scripts and npm
• Also offers a VS Code extension, though this README treats it as a separate product

Caveats

• The README never states which LLM backend powers the agents, whether API keys are required, or what running costs look like
• Only GitHub is mentioned; support for other git platforms or CI systems is unclear
• The relationship between the CLI checks and the VS Code extension is unexplained

Verdict

Teams who want lightweight, text-based AI review rules that live in git and surface directly in GitHub PRs should try this. If you need transparency on model choice, pricing, or non-GitHub support, wait for the docs to catch up.