How it works
crawwwl runs an audit in four stages: it gathers the right files, sends the audit for processing, analyzes the project, then returns a report to your terminal.
- 1
crawwwl gathers project files
The CLI scans your project, finds relevant source files, and collects useful context such as README.md, package.json, and tool instruction files.
- 2
crawwwl prepares the audit
The audit is sent to the crawwwl service and given an audit ID. The CLI keeps track of progress while the run is being processed.
- 3
crawwwl analyzes the project
crawwwl runs its scanner categories, groups repeated findings, calculates scores, and can use crawwwl AI to validate findings and add deeper analysis.
- 4
crawwwl returns the report
When the audit is complete, the CLI prints the report in your terminal. Use
--markdownto save a Markdown version alongside the terminal output.
How scoring works
Every audit gives your project one overall score, plus a separate score for each scanner category. Category scores run from 0 to 100, and findings reduce them based on severity, so higher-impact issues carry more weight.
The overall score is calculated from the categories that actually apply to the project. If a category does not apply, it is marked N/A and excluded so it does not pull the overall score down.
Excellent
Very few or no issues found. This category is in strong shape.
Good
Some issues are worth cleaning up, but nothing major is dragging the score down.
Needs work
Enough issues were found to justify a closer review and follow-up fixes.
Critical
Higher-impact issues are present and should be prioritized first.
Context files
crawwwl can read selected context files from your project root to better understand how the project is structured and how your team works in it.
CLAUDE.md # Claude Code project instructions
AGENTS.md # Agent/AI tool instructions
README.md # Project overview
tasks/todo.md # Task notes
package.json # Dependencies and scriptsThese files help crawwwl AI interpret findings in context. They do not suppress findings, change scores, or override scanner output. For more on how audit data is handled, see the privacy policy.
MCP integration
crawwwl includes an MCP (Model Context Protocol) server so you can run audits directly from supported AI coding tools without leaving your editor.
Supported tools include Claude Code, Cursor, and Windsurf and is also designed to work with other MCP-compatible tools, such as Codex.
MCP setup guide