claudekit

Real-time guardrails and workflow automation for Claude Code — catch errors instantly, save checkpoints, and delegate to expert subagents.

Requirements

Claude Code Max plan (recommended) · Node.js 20+ (tested on Node.js 24)

Install

npm install -g claudekit-dev
claudekit setup          # interactive setup wizard
claudekit setup --yes    # quick setup with defaults
claudekit setup --all    # install everything including all agents

What It Does

Claudekit is a safety net that runs alongside Claude Code:

| Without claudekit | With claudekit | |---|---| | Claude adds any type → discovered in code review | Instant block: "Use specific type: User \| null" | | Risky refactor fails → git archaeology | One command: /checkpoint:restore | | Claude breaks tests → you find out later | Claude sees the error immediately and fixes it | | Code review → shallow, sequential analysis | 6 specialized agents analyze in parallel |

Key Features

Real-time Error Prevention

Hooks run automatically as Claude works, catching issues before they compound:

• TypeScript guard — blocks any types and type errors on edit
• Linting — catches style issues immediately
• Test runner — runs relevant tests on file changes
• Security scanning — detects OWASP Top 10 vulnerabilities (hardcoded secrets, injection, weak crypto)
• Anti-patterns — prevents code replacement with comments, lazy refactoring
• Sensitive file protection — 195+ patterns across 12 categories (.env, SSH keys, cloud credentials)
• Package install guard — warns when installing packages without security scanning
• Multi-tool ignore support — .agentignore, .aiignore, .cursorignore, .geminiignore, etc.

Hooks reference → · File guard guide →

Git Checkpoint System

Undo any set of changes with one command:

• Auto-save — creates checkpoints when Claude stops
• Restore — /checkpoint:restore to undo changes
• Manage — /checkpoint:list, /checkpoint:create

Checkpoint guide →

Codebase Navigation

Claude sees your entire project structure automatically — no trial-and-error searches:

• Jump straight to files, functions, and classes
• Understand dependencies and architecture instantly
• Runs invisibly on first prompt, updates as you code
• Configurable include/exclude filters

Codebase map guide →

Multi-Agent Code Review

/code-review launches 6 specialized agents in parallel for architecture, security, performance, testing, quality, and documentation analysis. Results are prioritized with actionable suggestions.

Spec-Driven Development

6-phase iterative workflow: Implementation → Test Writing → Code Review → Improvement → Commit → Tracking. Dynamic agent selection with quality gates at each phase.

Spec workflow guide →

Expert Subagents

30+ specialized AI assistants for different domains:

| Category | Agents | |---|---| | Code Review | code-review-expert, triage-expert | | Security | security-expert | | TypeScript | typescript-expert, typescript-build-expert, typescript-type-expert | | React & Frontend | react-expert, react-performance-expert, nextjs-expert, css-styling-expert | | Testing | testing-expert, vitest-testing-expert, jest-testing-expert, playwright-expert | | Database | database-expert, postgres-expert, mongodb-expert | | Infrastructure | docker-expert, github-actions-expert, devops-expert | | Research | research-expert, code-search, oracle | | Frameworks | nestjs-expert, ai-sdk-expert, loopback-expert, kafka-expert |

claudekit list agents                      # list all with token counts
claudekit setup --agents typescript-expert,react-expert  # install specific agents

View all agents →

Slash Commands

# Git & Checkpoints
/checkpoint:create [msg]    Save current state
/checkpoint:restore [n]     Restore to checkpoint
/checkpoint:list            View all checkpoints
/git:commit                 Smart commit following conventions
/git:status                 Intelligent git analysis
/git:checkout [branch]      Smart branch creation/switching
/git:ignore-init            Initialize AI-safe .gitignore patterns

# Code Quality & Security
/code-review [target]       Multi-agent code review
/security-review [scope]    OWASP Top 10 security audit
/validate-and-fix           Run all quality checks and fix issues

# Development
/research [query]           Deep parallel research with subagents
/spec:create [feature]      Generate comprehensive specifications
/spec:execute [file]        Implement specs with iterative workflow
/create-subagent            Build custom AI assistants
/create-command             Create custom slash commands
/agents-md:init             Configure AGENTS.md for auto delegation

View all commands →

CLI Reference

# Setup & info
claudekit setup              # Interactive setup wizard
claudekit list               # Show all components
claudekit doctor             # Check installation health

# Upgrade & sync
claudekit upgrade            # Upgrade to latest version from npm
claudekit upgrade --check    # Check for updates without installing
claudekit sync               # Sync project components to current version
claudekit sync --dry-run     # Preview sync changes without modifying files
claudekit sync --project <path>  # Sync a specific project directory

# Extract prompts for external LLMs
claudekit show agent <id>    # Display agent prompt
claudekit show command <id>  # Display command prompt
claudekit show agent <id> -f json  # Output as JSON

# Hook management
claudekit-hooks run <hook>   # Execute a hook manually
claudekit-hooks profile      # Profile all hook performance
claudekit-hooks disable <h>  # Disable hook for current session
claudekit-hooks enable <h>   # Re-enable hook
claudekit-hooks status       # Show hook status

Hooks

Hooks automatically enforce quality as Claude works:

| Hook | Trigger | Description | |---|---|---| | file-guard | PreToolUse | Block access to sensitive files | | sfw-install | PreToolUse | Warn on unsecured package installs | | security-scan | PostToolUse | Scan for security anti-patterns | | typecheck-changed | PostToolUse | TypeScript checking on file changes | | lint-changed | PostToolUse | Linting on changed files | | test-changed | PostToolUse | Run tests for changed files | | check-any-changed | PostToolUse | Forbid any types | | check-comment-replacement | PostToolUse | Detect code→comment replacement | | check-unused-parameters | PostToolUse | Detect lazy _param refactoring | | codebase-map | UserPromptSubmit | Invisible codebase context | | thinking-level | UserPromptSubmit | Enhance Claude's reasoning | | create-checkpoint | Stop | Auto-checkpoint on stop | | self-review | Stop | Targeted self-review | | check-todos | Stop | Validate todo completions |

Hook configuration → · Profiling guide →

Configuration

.claude/settings.json — Project-level hook configuration:

{
  "hooks": {
    "PreToolUse": [{
      "matcher": "Read|Edit|MultiEdit|Write",
      "hooks": [{"type": "command", "command": "claudekit-hooks run file-guard"}]
    }],
    "PostToolUse": [{
      "matcher": "Write|Edit|MultiEdit",
      "hooks": [{"type": "command", "command": "claudekit-hooks run typecheck-changed"}]
    }],
    "Stop": [{
      "matcher": "*",
      "hooks": [{"type": "command", "command": "claudekit-hooks run create-checkpoint"}]
    }],
    "UserPromptSubmit": [{
      "matcher": "*",
      "hooks": [{"type": "command", "command": "claudekit-hooks run codebase-map"}]
    }]
  }
}

.claudekit/config.json — Hook-specific settings:

{
  "hooks": {
    "typecheck-changed": { "command": "npm run typecheck" },
    "thinking-level": { "level": 2 },
    "codebase-map": {
      "include": ["src/**", "lib/**"],
      "exclude": ["**/*.test.ts"],
      "format": "dsl"
    }
  }
}

Configuration guide →

Troubleshooting

claudekit doctor              # Check configuration and setup
claudekit-hooks status        # Check hook status
npm list -g claudekit-dev     # Verify installation

Full troubleshooting →

Development

git clone https://github.com/carlrannaberg/claudekit.git
cd claudekit && npm install
npm run build                 # Compile TypeScript
npm test                      # Run tests
npm run symlinks              # Create dev symlinks

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT