OpenClaw-CC

Autonomous AI Assistant Plugin for Claude Code

Turn Claude Code into a self-improving autonomous assistant with persistent memory, systematic debugging, automated releases, QA cycles, and multi-agent orchestration.

Quick Start · Features · Skills · Architecture · Plugin Install · Contributing

---

Why OpenClaw-CC?

Without	With OpenClaw-CC
Every session starts from zero	3-tier persistent memory across all sessions
Manual debugging guesswork	6-step systematic debugging with Iron Law
Copy-paste release workflow	8.5-step automated ship (test→review→PR)
No safety nets	Hook-based guardrails (freeze, careful, guard)
Single-agent limitations	19 OMC agents + 4 project agents orchestrated
No external notifications	Discord/Telegram real-time alerts
Stale API knowledge	4,400+ curated docs via Context Hub

---

Quick Start

Option A: Plugin Installation (Recommended)

# 1. Add marketplace
/plugin marketplace add https://github.com/Kit4Some/Oh-my-ClaudeClaw

# 2. Install plugin
/plugin install openclaw-cc@openclaw-cc

# 3. Install MCP dependencies
cd ~/.claude/plugins/cache/openclaw-cc/openclaw-cc/latest
cd mcp-servers/memory-manager && npm install && cd ../..
cd mcp-servers/knowledge-engine && npm install && cd ../..
cd mcp-servers/messenger-bot && npm install && cd ../..
cd mcp-servers/task-scheduler && npm install && cd ../..

Option B: Manual Installation

# 1. Clone
git clone https://github.com/Kit4Some/Oh-my-ClaudeClaw.git
cd Oh-my-ClaudeClaw

# 2. Install dependencies
for dir in mcp-servers/*/; do (cd "$dir" && npm install); done

# 3. (Optional) Install Context Hub for API doc fetching
npm install -g @aisuite/chub

# 4. Generate skills from templates
node scripts/gen-skill-docs.mjs

# 5. Launch
claude

Environment Setup (Optional)

For Discord/Telegram integration, create .env:

DISCORD_BOT_TOKEN=your_token
DISCORD_CHANNEL_ID=your_channel
DISCORD_WEBHOOK_URL=your_webhook
TELEGRAM_BOT_TOKEN=your_token
TELEGRAM_CHAT_ID=your_chat_id

Prerequisites

• Claude Code with active subscription
• oh-my-claudecode plugin
• Node.js >= 18

---

Features

3-Tier Persistent Memory

Episodic (30d)  →  Working (30d)  →  Long-term (permanent)
  daily-logs         tasks              knowledge
  captures           sessions           projects
                     inbox              people

• FTS5 full-text search with associative mode (5-signal ranking)
• Knowledge graph with 6 relation types (related, derived, supersedes, blocks, contradicts, refines)
• Trigram similarity search for deduplication (no external API)
• Automatic refinement pipeline: nightly dedup, weekly decay, monthly summary

Multi-Agent Orchestration

19 OMC agents + 4 project-local agents with intelligent routing:

Tier	Model	Agents
Quick	Haiku	explore, writer, comms-agent, session-manager
Standard	Sonnet	executor, debugger, tracer, verifier, test-engineer, designer, scientist, memory-specialist, research-agent
Complex	Opus	analyst, planner, architect, critic, code-reviewer, code-simplifier, product-manager

Safety System (from gstack)

Skill	Protection
/freeze	Blocks Edit/Write outside a chosen directory via PreToolUse hooks
/careful	Warns before rm -rf, DROP TABLE, force-push, reset --hard, etc.
/guard	Activates both freeze + careful simultaneously

Skill Template System

21 skills generated from .tmpl templates with 11 shared blocks — zero duplication:

node scripts/gen-skill-docs.mjs           # regenerate all
node scripts/gen-skill-docs.mjs ship       # regenerate one
node scripts/skill-check.mjs              # health dashboard

---

Architecture

┌──────────────────────────────────────────────────────────────┐
│                       Claude Code                            │
│                                                              │
│  ┌──────────┐  ┌───────────┐  ┌───────────────────────────┐ │
│  │ 21 Skills │  │ 4 Agents  │  │   oh-my-claudecode (OMC)  │ │
│  │ template  │  │ memory    │  │   19 specialized agents   │ │
│  │ generated │  │ comms     │  │   team orchestration      │ │
│  │           │  │ research  │  │   LSP/AST/Python tools    │ │
│  │           │  │ session   │  │   state management        │ │
│  └─────┬─────┘  └─────┬─────┘  └────────────┬──────────────┘ │
│        │               │                      │               │
│  ┌─────▼───────────────▼──────────────────────▼─────────────┐ │
│  │               5 MCP Servers (31 tools)                    │ │
│  │                                                           │ │
│  │  memory-manager (9)  ·  knowledge-engine (6)              │ │
│  │  messenger-bot  (4)  ·  task-scheduler  (7)               │ │
│  │  context-hub    (5)                                       │ │
│  └─────────────────────────┬─────────────────────────────────┘ │
│                            │                                   │
│  ┌─────────────────────────▼─────────────────────────────────┐ │
│  │              3-Tier Persistent Memory                      │ │
│  │                                                           │ │
│  │  Episodic (30d) ──→ Working (30d) ──→ Long-term (∞)      │ │
│  │                                                           │ │
│  │  SQLite FTS5 · Knowledge Graph · Trigram Similarity       │ │
│  │  Associative Search (5 signals) · Auto-Refinement         │ │
│  └───────────────────────────────────────────────────────────┘ │
│                            │                                   │
│  ┌─────────────────────────▼─────────────────────────────────┐ │
│  │           External Integrations                            │ │
│  │  Discord · Telegram · Context Hub (4,400+ API docs)       │ │
│  └───────────────────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘

---

MCP Servers (5 servers, 31 tools)

Server	Tools	Key Capabilities
memory-manager	9	memory_store memory_search memory_get memory_update memory_delete memory_daily_log memory_search_date memory_list memory_stats
knowledge-engine	6	memory_link memory_graph memory_similar memory_refine memory_archive memory_reindex_trigrams
messenger-bot	4	messenger_send messenger_read messenger_poll messenger_status
task-scheduler	7	task_create task_list task_update task_delete task_run_now task_history task_generate_crontab
context-hub	5	chub_search chub_get chub_list chub_annotate chub_feedback

---

Skills (21)

Workflow Skills

Skill	What It Does
/ship	8.5-step release automation: pre-flight → merge base → test → coverage audit → pre-landing review → version bump → changelog → bisectable commits → verification gate → push → PR → notify
/investigate	6-step systematic debugging: gather evidence → reproduce → scope lock → pattern analysis → hypothesis testing (3-strike rule) → verified fix with regression test
/code-review	Multi-pass review: scope drift detection → mechanical auto-fix (Pass 1) → security audit (Pass 2) → judgment items (Pass 3) → doc staleness check → WTF-likelihood gate
/qa	Test-Fix-Verify cycle: baseline → triage (Quick/Standard/Exhaustive) → fix loop with atomic commits → regression tests → WTF-likelihood self-regulation (hard cap: 50 fixes)
/office-hours	Idea validation: Startup mode (6 forcing questions) or Builder mode (generative brainstorming) → landscape search → premise challenge → forced alternatives → design document
/retro	Engineering retrospective: git + memory data → metrics table → time distribution → session detection → commit analysis → hotspots → focus score → streak tracking → trend comparison

Core Skills

Skill	What It Does
/task-analyzer	Decompose complex tasks → route to appropriate agents → execute → report
/memory-ops	Store, search, organize persistent memories with importance scoring
/research-collector	Multi-angle web research → structured output → dedup → store
/daily-routine	Morning briefing, task management, evening review, weekly retrospective
/doc-fetcher	Fetch API docs from Context Hub (4,400+ libraries) with annotations and feedback

Safety Skills

Skill	What It Does
/freeze	Block Edit/Write outside a chosen directory (hook-based, session-scoped)
/careful	Warn before destructive commands: rm -rf, DROP TABLE, force-push, reset --hard
/guard	Activate freeze + careful simultaneously for maximum safety
/unfreeze	Remove edit restrictions

Advanced Skills

Skill	What It Does
/knowledge-refiner	Detect duplicates, merge, archive stale, promote layers, reindex
/session-tracker	Dual-write to OMC state + persistent memory for cross-session continuity
/web-researcher	Multi-angle web search with evidence ranking and knowledge graph integration
/autonomous-ops	24/7 messenger-driven autonomous loop: poll → analyze → dispatch → persist → report
/knowledge-sync	Bidirectional sync between OMC ephemeral state and persistent memory
/deep-research	3 parallel research agents → analyst synthesis → critic review → knowledge graph

---

Builder Ethos

Five principles embedded in every skill via {{OCC_ETHOS}} template block (full document):

Principle	Core Idea
Boil the Lake	AI makes complete implementation near-free. Do the complete thing, every time.
Search Before Building	Memory (L0) → Standard patterns (L1) → Current trends (L2) → First principles (L3)
Build for Yourself	Specificity of a real problem beats generality of a hypothetical one
Memory is Cheap	Always store, always search. Let the refinement pipeline handle cleanup.
Delegate or Die	Right agent, right model. Never self-approve. Separate authoring from review.

---

Project Structure

Oh-my-ClaudeClaw/
├── .claude-plugin/
│   ├── plugin.json            # Plugin manifest
│   └── marketplace.json       # Marketplace catalog
├── .claude/
│   ├── agents/                # 4 project-local agents (OMC-quality prompts)
│   └── settings.local.json    # 5 lifecycle hooks
├── .mcp.json                  # 5 MCP server configurations
├── CLAUDE.md                  # Project instructions (loaded by Claude Code)
│
├── skills/                    # 21 skills
│   ├── {skill}/SKILL.md.tmpl  #   Template source (edit this)
│   └── {skill}/SKILL.md       #   Auto-generated (don't edit)
│
├── mcp-servers/
│   ├── memory-manager/        # 9 tools — SQLite FTS5 + associative search
│   ├── knowledge-engine/      # 6 tools — graph, similarity, refinement
│   ├── messenger-bot/         # 4 tools — Discord/Telegram
│   └── task-scheduler/        # 7 tools — cron + claude CLI execution
│
├── scripts/
│   ├── gen-skill-docs.mjs     # SKILL.md.tmpl → SKILL.md generator
│   ├── skill-check.mjs        # Health dashboard
│   ├── template-blocks/       # 11 shared blocks (preamble, memory, etc.)
│   └── hooks/                 # 5 lifecycle hooks (.mjs)
│
├── memory-store/              # Persistent memory (gitignored data)
├── docs/ETHOS.md              # Builder philosophy
├── package.json               # Build scripts
├── LICENSE                    # MIT
└── CONTRIBUTING.md            # Contribution guidelines

---

How It Works

User says: "debug this auth error"
     │
     ▼
┌─ Keyword Detection ──────────────────────────┐
│  Matches: "debug" → suggests /investigate     │
└───────────────────────────┬───────────────────┘
                            ▼
┌─ /investigate Skill ──────────────────────────┐
│  1. memory_search(tag: "bug") → past context  │
│  2. OMC tracer agent → evidence gathering     │
│  3. LSP tools → precise code navigation       │
│  4. Reproduce → Scope lock → Pattern match    │
│  5. 3-strike hypothesis testing               │
│  6. OMC executor → implement fix              │
│  7. OMC verifier → verify + regression test   │
│  8. memory_store → persist for future         │
│  9. messenger_send → notify completion        │
└───────────────────────────────────────────────┘

---

Translations

Language	Link
English	README.md (this file)
Korean (한국어)	README.ko.md
Chinese (中文)	README.zh.md
Japanese (日本語)	README.ja.md

---

Contributing

We welcome contributions! See CONTRIBUTING.md for:

• Development setup and workflow
• How to add new skills, MCP tools, or agents
• Template system conventions
• Pull request checklist

---

Credits

Built on top of:

• Claude Code by Anthropic
• oh-my-claudecode by Yeachan Heo
• gstack patterns by Garry Tan
• Context Hub by Andrew Ng

License

MIT © Evan Lee (Kit4Some)