Guide
OpenClaw + Claude Code
Claude Code is Anthropic's CLI front-end for Claude. You can use it as your terminal interface for an OpenClaw container — drive the agent, editopenclaw.json, manage skills, and ship the result to Telegram or Discord. Here is the wiring.
Why Pair Claude Code With OpenClaw?
Claude Code is great at code-and-config: editing JSON, writing skills, running shell commands, reading logs. OpenClaw is great at agent deployment: gateway, channels, memory, model routing. Connecting them gives you a powerful workflow — you develop your OpenClaw agent from Claude Code, then OpenClaw serves the result to end users on chat platforms.
Step 1: Install Claude Code
Install from the Anthropic CLI distribution channels. Once installed, log in with your Anthropic account.
# Verify Claude Code is installed
claude --version
# Login (opens browser)
claude loginStep 2: Get an OpenClaw Container
Either self-host via the install guide or deploy a managed container on OpenClaw Launch. Once you have a container, note its workspace path (usually /home/node/.openclaw) and gateway URL.
Step 3: Point Claude Code at the OpenClaw Workspace
Open Claude Code from inside the OpenClaw workspace so the CLI's file edits land where the gateway watches:
# Self-hosted
cd ~/.openclaw && claude
# Managed (via SSH to the container)
ssh openclaw@your-container "cd .openclaw && claude"From here, Claude Code can read openclaw.json, edit channel configs, draft new skills under extensions/, and run openclaw CLI commands directly.
Step 4: Share Workspace + Memory
The OpenClaw workspace at ~/.openclaw contains: openclaw.json,credentials/ (channel tokens), memory/ (Qwen embeddings on paid plans), and extensions/ (installed skills). Claude Code can read and edit all of these — just keep in mind:
- Edits to
channels.*,agents.*,models.*hot-apply - Edits to
plugins.*orgateway.auth.*need a gateway restart - Don't commit
credentials/— it has live tokens
Step 5: Drive the Agent From Claude Code
Claude Code can call the OpenClaw CLI as a tool. Ask Claude to send a test message, tail logs, or restart the gateway:
# In Claude Code:
"Send a test message via the OpenClaw gateway and report the response."
# Claude Code runs:
openclaw chat send "test"
openclaw logs --since 30sComparison With Hermes Agent + Claude Code
Hermes Agent has a similar Claude Code integration — see the Hermes side guide. The difference is scope: with OpenClaw you ship the result to Telegram / Discord / WhatsApp; with Hermes you ship to a single chat session.
Skip the Setup
OpenClaw Launch exposes an SSH-able shell into your managed container so Claude Code works out of the box. Plans from $3/mo first month.