claw-to-claude

claw-to-claude

Migrate an OpenClaw multi-agent setup to Claude Code Agent Teams.

What This Does

claw-to-claude is a Claude Code plugin that reads your existing OpenClaw configuration and workspace files, then generates Claude Code Agent Teams definitions ready to use. It produces persistent memory files, per-agent MCP server configurations, shared skill files, and scheduled task definitions — converting your entire OpenClaw org hierarchy into a set of coordinating CC agents with routing instructions and teammate messaging enabled.

Prerequisites

• An existing OpenClaw setup (running instance or a snapshot directory)
• Claude Code v2.1.32 or later (Agent Teams requires this version)
• Agent Teams enabled in your Claude Code settings:

  { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": true }
  

Installation

First, add the marketplace. Then install the plugin.

claude plugin marketplace add tkhattar14/claw-to-claude
claude plugin install claw-to-claude@claw-to-claude

Or from inside Claude Code:

/plugin marketplace add tkhattar14/claw-to-claude
/plugin install claw-to-claude@claw-to-claude

Usage

Open Claude Code and run:

/claw-to-claude:migrate --path ./openclaw-snapshot

Or connect directly to your OpenClaw host:

/claw-to-claude:migrate --ssh user@host

If using --path, copy your OpenClaw workspace to your local machine first:

scp -r user@host:~/openclaw-snapshot ./

How It Works

Migration runs in three phases:

Phase 1 — Analyze Reads all agent definitions (IDENTITY, SOUL, AGENTS, USER files), registered skills (SKILL.md), MCP server configs, cron jobs, org hierarchy, and long-term memory stores. Produces a structured inventory of everything in your OpenClaw setup.

Phase 2 — Plan Generates a migration plan and displays it for your review before any files are written. You can inspect what will be created, renamed, or skipped, and abort if something looks wrong.

Phase 3 — Execute Creates agent definition files under ~/.claude/agents/, merges or installs shared skills, documents MCP server configuration needed in project/user settings, converts cron jobs to CC scheduled triggers, and seeds agent memory from your existing MEMORY.md files.

What Gets Migrated

OpenClaw Feature	CC Equivalent
Agent definitions (IDENTITY, SOUL, AGENTS, USER)	~/.claude/agents/{name}.md with persistent memory
Skills (SKILL.md)	Merged into agent body or ~/.claude/skills/
MCP servers	Project/user MCP settings (shared across teammates)
Cron jobs	CC scheduled triggers
Long-term memory (MEMORY.md)	~/.claude/agent-memory/{name}/
Org hierarchy and routing	Lead agent routing instructions
Inter-agent communication	Agent Teams teammate messaging

What Does Not Migrate

• Per-agent Telegram/WhatsApp bots — CC uses a single channel; consolidate to one bot manually
• Vector/semantic memory search — CC uses file-based memory; semantic search is not available
• Daily notes — Ephemeral by design; not carried over
• Always-on daemon mode — Use a persistent tmux session instead

After Migration

List your migrated agents:

/agents

Start using them:

Create an agent team with [agent-name, agent-name] to [task description]

Configure any MCP server credentials manually in your agent frontmatter or ~/.claude/settings.json — credentials are not copied during migration.

Uninstall

claude plugins uninstall claw-to-claude

Note: generated agent files at ~/.claude/agents/ are not removed automatically.

License

MIT