MCP Integration

crawwwl for Claude Code, Cursor, and Windsurf

Give your AI coding tool direct access to crawwwl. Audit your project, explore findings, and fix issues, all from a conversation.

Works with any MCP-compatible AI tool.

Setup

1. Install crawwwl

npm install -g crawwwl

Quick and standard audits are free with no account required.

2. Add the MCP to your project

crawwwl mcp add --scope project

Writes a .mcp.json entry to your project directory. Use --scope user to apply globally across all projects.

2.1 Codex

[mcp_servers.crawwwl]
command = "crawwwl"
args = ["mcp"]
tool_timeout_sec = 420
startup_timeout_sec = 20
env_vars = ["CRAWWWL_API_URL", "CRAWWWL_API_KEY"]

For Codex, add this to ~/.codex/config.toml.

Then restart Codex and run /mcp to confirm crawwwl is available.

3. Install the agent skills

npx skills add crawwwl/agent-skills

Adds /crawwwl:audit and /crawwwl:fix slash commands, plus a workflow skill that auto-activates whenever you ask your AI tool to audit or fix your project.

Then restart your AI tool (Claude Code, Cursor, or Windsurf).

For deep audits (optional)

crawwwl login

Deep audits use a full reasoning model See pricing.

Available tools

Once connected, your AI tool has access to four tools:

crawwwl_audit

Run a full audit on any project path. Returns scores, all issues, and AI insights.

crawwwl_get_issues

Filter issues from the last report by severity or category, with file paths, line numbers, and fix suggestions.

crawwwl_get_report

Load the last saved report without re-running a scan.

crawwwl_fix_check

Re-scan and diff against the last report. Shows what was resolved and what is new.

What to say

Your AI tool will pick up the MCP automatically. Just describe what you want:

"Audit this project with crawwwl, verify each finding is real, then fix them systematically"
"Use crawwwl to find all high severity security issues in this repo and fix them"
"Run a crawwwl audit, then use crawwwl_fix_check after each fix to confirm progress"

How it works

  1. 1

    Your AI tool audits your project

    Your AI tool calls crawwwl_audit to scan your source files and get back a structured report: scores, issues, file paths, and line numbers.

  2. 2

    Your AI tool explores and confirms

    For each finding, your AI tool reads the flagged file at the exact line. It verifies the issue is real before attempting a fix. No blind edits.

  3. 3

    Your AI tool fixes systematically

    Issues are fixed one by one, in severity order. Each fix uses the recommendation and fix snippet directly from the crawwwl report.

  4. 4

    Your AI tool verifies with fix-check

    After applying fixes, your AI tool calls crawwwl_fix_check to re-scan and confirm the issues are resolved.

More

Documentationcrawwwl quick start, what it checks, and how it worksCLI referenceAll commands, flags, and output formatsPricingFree audits and deep audit pricing