n8n-nodes-claude-code-cli

---

✨ Features

🐳 Docker Execution - Run Claude Code in isolated containers	🔄 Session Management - Multi-turn conversations across executions
🎯 Tool Permissions - Fine-grained control over allowed tools	📁 Context Files - Include files and directories for analysis
🧠 Multiple Models - Opus, Sonnet, Haiku support	📊 Rich Output - Costs, tokens, and session IDs

---

⚡ Quick Start

1. Install the n8n node

In n8n: Settings > Community Nodes > Install > n8n-nodes-claude-code-cli

2. Deploy Claude Code Runner

n8n installed on host (not in Docker)? Use the standalone setup:

mkdir -p claude-code-runner && cd claude-code-runner && \
curl -fsSL https://raw.githubusercontent.com/ThomasTartrau/n8n-nodes-claude-code-cli/main/docker/production/claude-code/docker-compose.yml -o docker-compose.yml && \
curl -fsSL https://raw.githubusercontent.com/ThomasTartrau/n8n-nodes-claude-code-cli/main/docker/production/claude-code/Dockerfile -o Dockerfile && \
docker compose up -d --build

n8n also running in Docker? Use the complete stack instead

The node uses docker exec to communicate with claude-code-runner. This requires:

1. Docker CLI installed inside the n8n container
2. Docker socket mounted to access the Docker daemon

The standard n8nio/n8n image doesn't include Docker CLI, so we provide a custom setup:

mkdir -p n8n-claude-code && cd n8n-claude-code && \
curl -fsSL https://raw.githubusercontent.com/ThomasTartrau/n8n-nodes-claude-code-cli/main/docker/production/n8n-with-claude-code/docker-compose.yml -o docker-compose.yml && \
curl -fsSL https://raw.githubusercontent.com/ThomasTartrau/n8n-nodes-claude-code-cli/main/docker/production/n8n-with-claude-code/Dockerfile.n8n -o Dockerfile.n8n && \
curl -fsSL https://raw.githubusercontent.com/ThomasTartrau/n8n-nodes-claude-code-cli/main/docker/production/n8n-with-claude-code/Dockerfile.claude-code -o Dockerfile.claude-code && \
docker compose up -d --build

This builds a custom n8n image with Docker CLI and deploys both n8n and claude-code-runner together. Access n8n at http://localhost:5678

Already have n8n running in Docker? You can modify your existing setup:

1. Create a custom Dockerfile for n8n:

FROM docker:29-cli AS docker-cli
FROM n8nio/n8n
USER root
COPY --from=docker-cli /usr/local/bin/docker /usr/local/bin/docker
RUN chmod +x /usr/local/bin/docker
USER node

1. Update your docker-compose.yml:

services:
  n8n:
    build: .  # Use the custom Dockerfile above
    user: root  # Required for Docker socket access
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
      # ... your other volumes

1. Deploy claude-code-runner separately using the standalone setup above

Important: Both containers must be managed by the same Docker daemon.

3. Authenticate

docker exec -it claude-code-runner claude login

Follow the browser prompts to complete authentication.

4. Configure n8n credentials

Parameter	Value
Connection Type	Docker
Container Name	claude-code-runner
Working Directory	/workspace

5. Start automating 🚀

Search "Claude Code" in n8n node panel and create your first workflow.

---

🐳 Configuration

Workspace Setup

You have two options to work with your code:

Option 1: Clone repos inside container (recommended for isolation)

docker exec -it claude-code-runner git clone <repo-url>
# Or use git worktree for multiple branches

Option 2: Mount existing projects

volumes:
  - /path/to/your/project:/workspace/project-name

MCP Servers

Mount your MCP configuration to enable additional tools:

volumes:
  # MCP servers directory
  - ./mcp-servers:/root/.mcp
  # Or mount your local .mcp.json (avoids versioning credentials)
  - ~/.mcp.json:/root/.mcp.json:ro

🔐 Alternative: SSH deployment

For dedicated VM deployments (AWS EC2, GCP, etc.):

curl -fsSL https://deb.nodesource.com/setup_24.x | sudo -E bash -
sudo apt-get install -y nodejs git
npm install -g @anthropic-ai/claude-code
claude login

n8n Credentials:

Parameter	Value
Connection Type	SSH
Host	Your VM IP
Port	22
Auth Method	privateKey

---

⚙️ Node Operations

Operation	Description	Use Case
Execute Prompt	Send a prompt and get a response	Direct AI interaction
Execute with Context	Include files as context	Code review, analysis
Continue Session	Continue last conversation	Multi-turn interactions
Resume Session	Resume specific session by ID	Continue after interruption

📝 Detailed Parameters

Execute Prompt

• Prompt (required): Instruction for Claude Code
• Model: Claude model to use
• Options: Working directory, timeout, system prompt

Execute with Context

• Prompt (required): Instruction for Claude Code
• Context Files: File paths to include
• Additional Directories: Directory paths

Continue / Resume Session

• Prompt (required): Follow-up message
• Session ID (resume only): Previous session ID

---

💡 Use Cases

🔍 MR/PR Code Review Agent

┌─────────────┐     ┌──────────────┐     ┌─────────────┐     ┌──────────────┐
│   GitLab    │────▶│  Get Diff &  │────▶│ Claude Code │────▶│    Post      │
│   Webhook   │     │    Files     │     │   Review    │     │   Comments   │
└─────────────┘     └──────────────┘     └─────────────┘     └──────────────┘

Workflow Steps

1. Trigger: GitLab/GitHub webhook on new MR/PR
2. Fetch: Get changed files and diff via API
3. Review: Claude Code with executeWithContext
   • "Review this code. Check for bugs, security issues, suggest improvements."
4. Post: Send review comments back to GitLab/GitHub
5. Notify: Alert team via Slack/Discord (optional)

---

🎧 Support Assistant

┌─────────────┐     ┌──────────────┐     ┌─────────────┐     ┌──────────────┐
│   Support   │────▶│   Analyze    │────▶│ Claude Code │────▶│   Respond    │
│   Ticket    │     │    Issue     │     │   Solution  │     │   or Route   │
└─────────────┘     └──────────────┘     └─────────────┘     └──────────────┘

Workflow Steps

1. Trigger: Webhook from support system (Zendesk, Intercom)
2. Analyze: Claude Code understands the issue
   • "User reports: [issue]. Analyze and suggest solution."
3. Respond: Send AI response back via API
4. Escalate: Route complex issues to humans

---

📚 Auto Documentation

┌─────────────┐     ┌──────────────┐     ┌─────────────┐     ┌──────────────┐
│    Code     │────▶│  Get Changed │────▶│ Claude Code │────▶│   Commit     │
│    Push     │     │    Files     │     │  Gen Docs   │     │    Docs      │
└─────────────┘     └──────────────┘     └─────────────┘     └──────────────┘

Workflow Steps

1. Trigger: Webhook on code push to main
2. Identify: Get list of changed files
3. Generate: Claude Code generates documentation
   • "Generate docs for this code. Include descriptions, params, examples."
4. Commit: Create commit with updated docs
5. PR: Optionally create a PR for review

---

🐛 Auto Bug Fixing

┌─────────────┐     ┌──────────────┐     ┌─────────────┐     ┌──────────────┐
│   Sentry    │────▶│    Parse     │────▶│ Claude Code │────▶│  Create PR   │
│   Alert     │     │ Stack Trace  │     │   Fix Bug   │     │   + Notify   │
└─────────────┘     └──────────────┘     └─────────────┘     └──────────────┘

Workflow Steps

1. Trigger: Webhook from error monitoring (Sentry, Datadog)
2. Analyze: Parse error stack trace
3. Fix: Claude Code analyzes and fixes
   • "Error: [stack trace]. Analyze code and provide a fix."
4. Test: Run tests to validate
5. PR: Create pull request with fix
6. Notify: Alert team about automated fix

---

🤖 Cloud Coding Bots

Build AI coding assistants on Telegram, Slack, Discord, or GitLab/GitHub.

Examples

📱 Telegram Bot

1. Trigger: Telegram trigger on new message
2. Process: Claude Code handles coding question
3. Reply: Send response via Telegram node

💬 Slack Bot

1. Trigger: Slack mention or slash command
2. Context: Fetch relevant code from repos (optional)
3. Respond: Post response to channel

🦊 GitLab/GitHub Bot

1. Trigger: Issue comment with keyword (e.g., /claude)
2. Analyze: Fetch issue context and code
3. Comment: Post Claude's analysis

---

📤 Output Structure

{
  "success": true,
  "sessionId": "550e8400-e29b-41d4-a716-446655440000",
  "output": "Here's my analysis of the code...",
  "exitCode": 0,
  "duration": 15234,
  "cost": 0.0523,
  "numTurns": 3,
  "usage": {
    "inputTokens": 1250,
    "outputTokens": 890
  }
}

Field	Description
success	✅ Execution completed successfully
sessionId	🔗 ID for continuing conversations
output	📝 Response text from Claude Code
cost	💰 Estimated cost in USD
usage	📊 Token breakdown

---

🔒 Security

🛡️ Tool Permissions

Control what Claude Code can do:

✅ Allowed: Read, Glob, Grep
❌ Blocked: Bash(rm:*), Write(.env)

🐳 Isolation Best Practices

• Always set specific working directory
• Avoid / or home directories
• Create dedicated workspace per project

🔐 Recommended Security Settings

Disallow dangerous operations:

• Bash(rm:*) - Prevent file deletion
• Bash(sudo:*) - No sudo access
• Write(.env) - Protect secrets
• Bash(curl:*) - Block network (if not needed)

---

🤝 Contributing

Development Setup

git clone https://github.com/ThomasTartrau/n8n-nodes-claude-code-cli.git
cd n8n-nodes-claude-code-cli
npm install && npm run build

# Start n8n + claude-code-runner for testing
docker compose -f docker/development/docker-compose.yml up -d --build

# Authenticate claude-code-runner
docker exec -it claude-code-runner claude login

# Access n8n at http://localhost:5678

Docker Files Structure

docker/
├── development/
│   ├── Dockerfile          # n8n with docker CLI (dev)
│   └── docker-compose.yml  # n8n + claude-code-runner (mounts dist/)
└── production/
    ├── claude-code/
    │   ├── Dockerfile          # Standalone claude-code-runner
    │   └── docker-compose.yml  # For n8n on host (not Docker)
    └── n8n-with-claude-code/
        ├── Dockerfile.n8n          # n8n with Docker CLI
        ├── Dockerfile.claude-code  # claude-code-runner
        └── docker-compose.yml      # Complete stack (both services)

Submit Changes

git checkout -b feature/amazing-feature
git commit -m 'feat: add amazing feature'
git push origin feature/amazing-feature

---

📄 License

MIT License - see LICENSE for details.

---