CodeMind v2.0 Documentation

Activating via "use codemind"

In any MCP-compatible IDE (Cursor, Windsurf, Claude Code), simply type:

"use codemind"

This activates the full Guardian v2.0 suite: SAST scanning, secrets detection, SCA, IaC auditing, documentation fetching, and auto-fix.

All MCP Tools (14)

🔍 Static Analysis (SAST)

• guard_code

  Comprehensive security & quality audit — 50+ patterns for SQLi, XSS, SSRF, command injection, path traversal, credential exposure, and AI slop detection. Returns a scored report.

• scan_and_fix

  Detects vulnerabilities and auto-fixes them using LLM. Returns original issues, fixed code, and a diff.

• improve_code

  Automatically refactors code to fix security and quality issues found by the auditor.

🛡️ Deep Security Scan

• deep_security_scan

  Multi-layer analysis combining SAST + Secrets + Quality in one call. Runs all engines and returns a unified security score, severity breakdown, and formatted report.

🔑 Secrets Detection

• scan_secrets

  Deep secrets detection with 30+ API key patterns (AWS, GCP, Stripe, GitHub, etc.) plus Shannon entropy analysis to catch unknown/custom secrets. Minimizes false positives via file-type filtering.

📦 Software Composition Analysis (SCA)

• scan_dependencies

  Scans lockfiles (requirements.txt, package-lock.json, yarn.lock, go.sum, Cargo.lock, etc.) and checks against Google's OSV.dev vulnerability database. Privacy-preserving: only package names & versions are sent.

• check_package

  Quick single-package CVE lookup. Pass a name, version, and ecosystem (PyPI, npm, Go, etc.) to check for known vulnerabilities.

🏗️ Infrastructure as Code (IaC)

• scan_iac_file

  Scans Dockerfiles, GitHub Actions workflows, and docker-compose files for security misconfigurations — root user, secrets in ENV, unpinned images, supply chain attacks, excessive permissions.

• scan_infrastructure

  Scans ALL IaC files in a project directory automatically. Finds and audits Dockerfiles, CI workflows, and compose files.

📋 Reporting

• export_security_report

  Exports scan results as SARIF v2.1.0 (GitHub Code Scanning compatible), JSON, Markdown, HTML, or CSV. Includes CWE/OWASP mappings and code snippets.

📚 Documentation

• resolve_library

  Resolve a library name to a documentation ID (like Context7).

• query_docs

  Fetch up-to-date documentation for any library with code examples.

• detect_code_libraries

  Detect frameworks and libraries in code snippets and suggest documentation queries.

🔧 Git & Review

• review_diff

  Generate an AI code review prompt from git diff changes.

• validate_ai_response

  Validate AI review output format and check for critical issues.

• get_review_history

  Get recent code review history and statistics.

• get_git_context

  Get current git repository context (branch, upstream, state).

📖 Resources & Prompts

• guardian://best-practices

  Resource with authoritative security & clean code best practices.

• code_review_prompt

  Generate a guided prompt for AI code review (general, security, performance, refactor).

• secure_implementation_prompt

  Generate a prompt to implement code with security in mind.

• codemind

  The unified "use codemind" prompt that orchestrates the full workflow.

Installation

Install CodeMind with MCP support using pip:

pip install "codemind[mcp]"

How It Works

1. Server Start

   The MCP server runs locally, exposing security and quality tools to your LLM.

2. AI Interaction

   When you say "use codemind", the AI receives instructions to use our Guardian tools.

3. Local Audit

   Your code stays 100% local. The audit happens on your machine, never in the cloud.

codemind install

Install the git pre-push hook in the current repository.

codemind install [--force]

codemind uninstall

Remove the git pre-push hook from the current repository.

codemind uninstall

codemind run

Manually run AI code review on current changes without pushing.

codemind run [OPTIONS]

Interactive Mode

When running with -i, CodeMind will prompt you to paste the AI's response. It then parses the feedback and lets you interactively review each issue.

codemind commit

Generate professional AI commit messages from staged changes.

codemind commit [--style STYLE] [--apply]

codemind rules

Manage and run custom review rules.

codemind rules list [--preset PRESET]
codemind rules check [-p PRESET] [-f FILE]

codemind ci

Integrate CodeMind into your CI/CD pipeline.

codemind ci init [--output PATH]
codemind ci info

Automatically generates a GitHub Actions workflow that runs CodeMind on every pull request.

codemind pr create

Generate high-quality Pull Request descriptions from branch changes.

codemind pr create [--base BRANCH] [--copy]

Automates the creation of professional summaries including Overview, Key Changes, and Impact.

codemind gateway

Manage the CodeMind REST API Gateway.

codemind gateway start [--host HOST] [--port PORT] [--reload]

Starts the FastAPI server to expose CodeMind functionality over REST.

codemind notify

Send review summaries to Slack or Discord.

codemind notify slack WEBHOOK_URL
codemind notify discord WEBHOOK_URL

Pushes latest review results to team communication channels.

codemind status

Show CodeMind status including config, hook, and detected IDE.

codemind status

Output includes:

• Config file location (or "using defaults")
• Hook installation status
• Currently detected IDE

codemind config

Manage configuration files and team sharing.

codemind config init [--wizard]
codemind config show
codemind config export [--output PATH]
codemind config import SOURCE

Interactive Wizard

Run codemind config init --wizard for a guided 30-second setup experience including IDE detection and hook installation.

Team Sharing

Sync your review standards across the team by exporting your config and sharing the file, or a URL to the config.

codemind doctor

Run comprehensive system health and diagnostic checks.

codemind doctor

Checks Git repo health, hook integrity, config validity, IDE connectivity, and Standalone LLM status.

codemind template

Manage and customize prompt templates.

codemind template [OPTIONS]

Template Placeholders

• {max_comments} - Maximum number of comments
• {branch_name} - Current branch name
• {file_count} - Number of files changed
• {additions} - Lines added
• {deletions} - Lines deleted
• {diff_content} - The actual diff

codemind detect

Detect running IDEs and show their details.

codemind detect

Shows all detected IDEs with:

• IDE name and version
• Window title
• AI chat shortcut key

codemind history

View review history and statistics.

codemind history [OPTIONS]

codemind serve

Run CodeMind as an MCP (Model Context Protocol) server.

codemind serve [OPTIONS]

Available MCP Tools (14)

• 🔍 guard_code — SAST audit (50+ patterns)
• 🔧 scan_and_fix — Detect + auto-fix vulnerabilities
• ✨ improve_code — Refactor and improve code
• 🛡️ deep_security_scan — Multi-layer security analysis
• 🔑 scan_secrets — 30+ API key patterns + entropy
• 📦 scan_dependencies — SCA via OSV.dev
• 🔍 check_package — Single package CVE check
• 🏗️ scan_iac_file — Dockerfile/Actions/Compose security
• 🏗️ scan_infrastructure — Full project IaC scan
• 📋 export_security_report — SARIF/JSON/MD/HTML/CSV
• 📚 resolve_library — Library ID resolution
• 📖 query_docs — Documentation fetching
• 🔎 detect_code_libraries — Framework detection
• 📝 review_diff — Git diff review prompt

Config File

Create a .codemind.yml file in your project root:

# When to run reviews: always, ask, never
enabled: always

ide:
  # Preferred IDEs (in order)
  preferred:
    - cursor
    - claude-code
    - windsurf
    - vscode
  # Auto-inject prompt into IDE
  auto_inject: true
  # Auto-submit the prompt
  auto_submit: false

review:
  # Maximum comments per review
  max_comments: 10
  # Enforce strict JSON output format
  strict_format: true

prompt:
  # Custom template file (optional)
  template_path: null
  # Extra rules to include
  extra_rules: []

Options Reference

Custom Templates

Create custom prompt templates for your team:

# Export the default template
codemind template --export -o my-template.txt

# Edit my-template.txt with your customizations

# Reference in config
prompt:
  template_path: my-template.txt

Cursor

Status: Full Support

CodeMind auto-injects prompts via Ctrl+L into Cursor's AI chat.

Claude Code

Status: Full Support

Native integration with Claude's AI assistant.

Windsurf

Status: Full Support

Works with Codeium's Cascade AI.

VS Code

Status: Copilot Required

Works with GitHub Copilot Chat extension. You can configure CodeMind as an MCP server in your settings.json or settings.json:

{
  "mcp": {
    "servers": {
      "CodeMind": {
        "command": "codemind",
        "args": ["serve"]
      }
    }
  },
  "github.copilot.chat.codeGeneration.instructions": [
    {
      "text": "When generating code, use CodeMind to audit for security and quality."
    }
  ]
}