mirror of
https://github.com/agentscope-ai/ReMe.git
synced 2026-10-28 06:24:43 +00:00
93a82deb3e
Closes #347 ## What - New OpenClaw plugin at plugins/openclaw/ following the same architecture as Claude Code / Codex plugins - .openclaw-plugin/plugin.json manifest - Stop hook (hooks/auto_memory.py) with detached daemon pattern, using OPENCLAW_PLUGIN_ROOT env var - Memory recall skill (skills/reme-memory/SKILL.md) adapted for OpenClaw's native MCP management - Bundled .mcp.json for HTTP MCP connection, documented openclaw mcp set/add workflow - Full README with install, prerequisites, structure, and troubleshooting ## Why - ReMe already ships plugins for Claude Code and Codex. OpenClaw was requested in #347 with a help-wanted label, and the contributor-provided design note aligned with the existing plugin architecture (manifest, Stop hook for auto_memory_cc, recall skill, MCP-configured via the host natively). - Keeping OpenClaw aligned with the other plugins makes behavior and maintenance predictable. ## Not in scope - No vector-database or storage backend work (covered by #386) - No change to the MCP boundary, core steps, or server-side agents ## Testing - [x] Pre-commit on all new files: check-json, ast, black, flake8, pylint, pyroma, add-trailing-comma, whitespace, private-key — all PASS - [x] auto_memory.py validated syntactically (python -m py_compile) - [x] JSON files parse cleanly (no trailing commas) - Manual OpenClaw integration steps are documented in README: the plugin ships .mcp.json so users can point openclaw mcp set at it ## Impact - Breaking change: none - Affected modules: new directory plugins/openclaw/ (no existing code touched) - Docs: README.md under plugins/openclaw/ |
||
|---|---|---|
| .. | ||
| claude_code | ||
| hermes_agent | ||
| openclaw | ||