Skillget

openclaw-claude-code

mcp
SUMMARY

OpenClaw plugin — turn Claude Code CLI into a programmable, headless coding engine with 14 tools, agent teams, and multi-model proxy

README.md

openclaw-claude-code

Full-featured Claude Code integration for OpenClaw — session management, agent teams, multi-model proxy, and plan mode workflows.

npm version
CI
License: MIT

What is this?

An OpenClaw native plugin that turns Anthropic's Claude Code CLI into a programmable, headless coding engine. Your AI agents get 10 tools to drive Claude Code sessions — start, send messages, manage context, coordinate agent teams, and more.

Works as:

  • OpenClaw Plugin — install once, agents get claude_session_* tools automatically
  • Standalone CLI — claude-code-skill serve + CLI commands, no OpenClaw needed
  • TypeScript library — import { SessionManager } from '@enderfga/openclaw-claude-code'

Install

As OpenClaw Plugin

openclaw plugins install @enderfga/openclaw-claude-code
openclaw gateway restart

That's it. Your agents can now use claude_session_start, claude_session_send, etc.

Standalone (no OpenClaw)

npm install -g @enderfga/openclaw-claude-code

# Start the embedded server
claude-code-skill serve

# Use CLI commands
claude-code-skill session-start myproject -d ~/project
claude-code-skill session-send myproject "fix the auth bug"
claude-code-skill session-stop myproject

Tools (14)

Tool Description
claude_session_start Start a session with full CLI flag support (model, effort, worktree, bare, agent teams, etc.)
claude_session_send Send a message and get the response
claude_session_stop Stop a session
claude_session_list List active sessions (also returns persisted sessions array)
claude_session_status Status with context %, tokens, cost, uptime
claude_session_grep Search session history by regex
claude_session_compact Compact session to reclaim context window
claude_session_update_tools Update allowed/disallowed tools at runtime (restarts with --resume)
claude_session_switch_model Switch model for a running session (restarts with --resume)
claude_agents_list List agent definitions from .claude/agents/
claude_team_list List teammates in an agent team session
claude_team_send Send message to a specific teammate
claude_session_health Health check for a specific session
claude_sessions_overview Plugin health overview: all sessions, stats, version

CLI Commands

Session Management

claude-code-skill session-start [name] [options]
claude-code-skill session-send <name> <message> [--effort high] [--plan]
claude-code-skill session-stop <name>
claude-code-skill session-list
claude-code-skill session-status <name>
claude-code-skill session-grep <name> <pattern>
claude-code-skill session-compact <name> [--summary <text>]

Agent / Skill / Rule Management

claude-code-skill agents-list [-d <dir>]
claude-code-skill agents-create <name> [--description <desc>] [--prompt <prompt>]
claude-code-skill skills-list [-d <dir>]
claude-code-skill skills-create <name> [--description <desc>] [--prompt <prompt>]
claude-code-skill rules-list [-d <dir>]
claude-code-skill rules-create <name> [--paths "*.py"] [--condition "Bash(git *)"]

Agent Teams

claude-code-skill session-start team -d ~/project --enable-agent-teams
claude-code-skill session-team-list <name>
claude-code-skill session-team-send <name> <teammate> <message>

Session Start Flags

All Claude Code CLI flags are supported:

claude-code-skill session-start myproject \
  -d ~/project \
  -m opus \
  --effort high \
  --bare \
  --worktree \
  --fallback-model sonnet \
  --json-schema '{"type":"object","properties":{"name":{"type":"string"}}}' \
  --mcp-config ./mcp.json \
  --settings ./settings.json \
  --skip-persistence \
  --betas "max-tokens-3-5-sonnet-2024-07-15" \
  --enable-agent-teams \
  --allowed-tools Bash,Read,Edit \
  --max-turns 50 \
  --max-budget 5.00

Multi-Model Proxy

Built-in Anthropic-to-OpenAI format translation. Claude Code CLI talks Anthropic format; the proxy converts to/from OpenAI format for Gemini, GPT, and other models.

  • Pure TypeScript, zero Python dependency
  • Streaming SSE conversion
  • Gemini tool schema cleaning
  • Thought signature caching (Gemini round-trip)
  • Gateway passthrough mode (OpenClaw handles routing)

Architecture

openclaw-claude-code/
├── src/
│   ├── index.ts                 # Plugin entry — 10 tools + hooks + proxy route
│   ├── types.ts                 # Shared types, model pricing/aliases
│   ├── persistent-session.ts    # Claude CLI subprocess management
│   ├── session-manager.ts       # Pure class, manages multiple sessions
│   ├── embedded-server.ts       # Auto-start HTTP server for CLI
│   ├── hooks/
│   │   └── prompt-bypass.ts     # Passthrough workspace hook
│   └── proxy/
│       ├── handler.ts           # HTTP route handler
│       ├── anthropic-adapter.ts # Anthropic ↔ OpenAI format conversion
│       ├── schema-cleaner.ts    # Gemini schema compatibility
│       └── thought-cache.ts     # Gemini thought signature cache
├── bin/
│   └── cli.ts                   # CLI (HTTP client to embedded server)
├── skills/
│   └── SKILL.md                 # Bundled skill for plan mode workflows
├── openclaw.plugin.json         # Plugin manifest
└── package.json

Configuration (OpenClaw)

In ~/.openclaw/openclaw.json:

{
  "plugins": {
    "entries": {
      "openclaw-claude-code": {
        "enabled": true,
        "config": {
          "claudeBin": "claude",
          "defaultModel": "claude-opus-4-6",
          "defaultPermissionMode": "acceptEdits",
          "defaultEffort": "auto",
          "maxConcurrentSessions": 5,
          "sessionTtlMinutes": 120
        }
      }
    }
  }
}

Requirements

  • Node.js >= 22
  • Claude Code CLI installed (npm install -g @anthropic-ai/claude-code)
  • OpenClaw >= 2026.3.0 (for plugin mode, optional)

License

MIT

Yorumlar (0)

Sonuc bulunamadi