Setting Up Claude Code with OpenClaw: Step-by-Step Guide

Why Claude Code + OpenClaw?

OpenClaw is a gateway for AI agents that gives you a persistent workspace, tool access, and multi-agent orchestration. When paired with Claude Code as a coding agent, you get a powerful development environment where Claude Code can work autonomously on complex tasks while OpenClaw manages context, tools, and delivery.

What This Integration Enables

- Background coding tasks — Claude Code runs while you continue working - Persistent workspace — OpenClaw maintains project context across sessions - Tool orchestration — Claude Code gets file system, search, and automation access - Multi-agent workflows — Claude Code + other agents working in parallel - Scheduled tasks — Code review, build, and deployment on schedule

Installation

Step 1: Install OpenClaw

# Install OpenClaw via npm
npm install -g openclaw
# Start the gateway
openclaw gateway start
# Check status
openclaw status

Step 2: Install Claude Code

npm install -g @anthropic-ai/claude-code
# Verify installation
claude --version

Step 3: Set Your API Keys

Configure both tools with your API keys:

# Claude Code needs an Anthropic API key
export ANTHROPIC_API_KEY="sk-ant-xxxxxxxx"
# OpenClaw can use various providers
# Set the model provider in your config

Add these to your shell profile (~/.zshrc or ~/.bashrc) for persistence:

echo 'export ANTHROPIC_API_KEY="sk-ant-xxxxxxxx"' >> ~/.zshrc
source ~/.zshrc

Configuration

Configure OpenClaw for Claude Code

Create or edit ~/.openclaw/config.yaml:

# OpenClaw configuration for Claude Code
gateway:
  port: 3001
  heartbeatInterval: 30
agents:
  claudeCode:
    command: claude
    args: ["--permission-mode", "bypassPermissions", "--print"]
    model: claude-sonnet-4-20250514
    timeout: 300

Create a Skills Configuration

OpenClaw uses "skills" to give agents specialized abilities. Create a skill for Claude Code:

mkdir -p ~/.openclaw/skills/coding-agent

Create ~/.openclaw/skills/coding-agent/SKILL.md:

---
name: coding-agent
description: Run Claude Code for autonomous coding tasks
---
# Coding Agent Skill
Use Claude Code for:
- Code generation and refactoring
- Bug fixing and debugging
- Test writing
- Code review
- Documentation generation

Usage Patterns

Pattern 1: One-Shot Coding Tasks

Run Claude Code through OpenClaw for specific tasks:

# OpenClaw spawns Claude Code as a sub-agent
openclaw run \
  --agent claudeCode \
  --task "Add a dark mode toggle to the React app. 
          Update the components, styles, and add local storage persistence."

Claude Code will analyze the project, implement the changes, and report back.

Pattern 2: Scheduled Code Review

Set up automatic daily code review:

# In OpenClaw cron configuration
schedule:
  - time: "09:00"
    agent: claudeCode
    task: "Review all PRs opened yesterday in the current repo.
          Check for: security issues, code style violations, 
          missing tests, and performance concerns."

Pattern 3: Multi-Agent Workflow

Combine Claude Code with other agents for complex workflows:

# Build pipeline with multiple agents
workflow:
  steps:
    - agent: claudeCode
      task: "Write unit tests for all new API endpoints"
    
    - agent: claudeCode
      task: "Review the tests for completeness and edge cases"
    
    - agent: claudeCode
      task: "Fix any failing tests and optimize test coverage"

Pattern 4: Project Scaffolding

Use Claude Code through OpenClaw to set up new projects:

openclaw run \
  --agent claudeCode \
  --task "Create a new Next.js project with TypeScript, 
          Tailwind CSS, Prisma ORM, and authentication.
          Initialize the project structure and install dependencies."

Real-World Example: Bug Fix Workflow

Here's how a real bug fix flows through the system:

1. Developer: "Fix the login redirect bug"
   
2. OpenClaw sends the task to Claude Code
   
3. Claude Code:
   - Reads the auth components
   - Identifies the redirect logic issue
   - Fixes the redirect URL construction
   - Adds a test for the edge case
   - Commits the changes
   
4. OpenClaw reports completion with:
   - Files changed
   - What the bug was
   - How it was fixed
   - Test results

Monitor Progress in Real-Time

# Check active agent sessions
openclaw status
# View recent agent output
openclaw logs --agent claudeCode --lines 20

Performance Tips

1. Set Cost Limits

Claude Code API costs can add up. Configure limits:

# Claude Code cost controls
claude config set maxCost 20
claude config set costThreshold 15
# Warn when spending exceeds threshold
claude config set costNotification true

2. Optimize Context Windows

For large projects, limit context to relevant areas:

# Focus Claude Code on specific directories
claude --context src/pages/ src/components/
# Or use .claudeignore to exclude files
echo "node_modules/\n.next/\ndist/" >> .claudeignore

3. Use MCP Servers

Extend Claude Code's capabilities through MCP servers managed by OpenClaw:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@anthropic-ai/mcp-filesystem"],
      "env": {
        "ALLOWED_PATHS": "/home/you/projects"
      }
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@anthropic-ai/mcp-github"],
      "env": {
        "GITHUB_TOKEN": "ghp_xxxx"
      }
    }
  }
}

Troubleshooting

Claude Code Not Responding

# Check if Claude Code works independently
claude --print "Hello, are you working?"
# Check OpenClaw gateway status
openclaw gateway status
# Restart if needed
openclaw gateway restart

Permission Issues

# Ensure Claude Code has project access
ls -la /path/to/your/project
# Claude Code might need read/write permissions on the project directory
chmod -R 755 /path/to/your/project

Timeout Errors

# Increase timeout in OpenClaw config
# config.yaml
agents:
  claudeCode:
    timeout: 600  # 10 minutes instead of 5

Is This Setup Right for You?

| You should use this setup if... | You're better off with just Claude Code if... | |--------------------------------|----------------------------------------------| | Managing multiple projects | Working on a single project | | Need scheduled/recurring tasks | Prefer interactive coding sessions | | Want multi-agent orchestration | Don't need complex automation | | Need persistent agent workspace | Run ad-hoc tasks only |

Summary

Claude Code + OpenClaw gives you an autonomous coding agent that runs on your terms — whether that's one-shot tasks, scheduled maintenance, or complex multi-agent workflows.

The setup takes about 15 minutes and scales from solo development to team-scale automation.

---

New to Claude Code? Start with our installation guide. Want to learn more about MCP? Check out our MCP beginner's guide.