Multi-agent personas. Dual-layer memory with contradiction detection. 23 LLM providers with auto-fallback. 208-pattern security scanner. 121 built-in skills, 60+ Citedy MCP tools. Apache 2.0.
No Docker needed. Local install for development and testing.
pip install adclaw
adclaw init
adclaw app
121 skills, Chromium, Playwright, all messaging channels (Telegram, Discord, Feishu, DingTalk).
docker run -d --name adclaw \
-p 8088:8088 \
nttylock/adclaw:latest
Or run our hosted version — $10/mo →
Apache 2.0 · v1.0.6 · 121 skills shipped · Active development
You're going to grep the source anyway. Here's the map.
Each persona has its own identity, LLM config, skill set, MCP tools, and schedule. Coordinator persona orchestrates multi-step campaigns.
Per-agent ReMe plus shared AOM with SQLite, sqlite-vec, FTS5, typed memory categories, boosted feedback retrieval, and contradiction detection.
Markdown cleanup, context tiers, near-deduplication, smart consolidation, and pruning keep long-running marketing context usable.
Static analysis and analysis-first LLM audit run before automation. Critical findings short-circuit; memory sanitizer catches prompt injection.
OpenAI, Anthropic, Gemini, DeepSeek, Groq, OpenRouter, Mistral, xAI, Ollama, vLLM, Azure OpenAI, and more with auto-fallback.
Marketing automation tools for SEO audits, content gaps, competitor research, lead magnets, video shorts, trend scouting, GA, and GSC.
| Channel | Status |
|---|---|
| Telegram | Primary |
| Discord | Built-in |
| Web Console | Built-in |
| iMessage | Built-in |
| DingTalk | Built-in |
| Feishu | Built-in |
| Built-in |
@tag routing across all channels. Coordinator-aware. Shared memory.
| Capability | AdClaw | CoPaw | CrewAI | Dify | OpenClaw MC |
|---|---|---|---|---|---|
| LLM providers | 23 (100+ models) | ~5 | Custom | ~10 | Custom |
| LLM auto-fallback | Chain | No | No | No | No |
| Multi-agent personas | Yes | No | Yes | No | Yes |
| Built-in skills | 121 | ~50 | Custom | Plugins | — |
| Marketing tools (MCP) | 52 | No | No | No | No |
| Chat channels | 7 | 5 | API | API + Web | API |
| Web dashboard | Yes | Basic | No | Yes | Yes |
| Per-persona chat tabs | Yes | No | No | No | No |
| Memory system | ReMe + AOM | ReMe | Short-term | RAG | — |
| Memory optimization (R1-R5) | 5 layers + contradictions | No | No | No | No |
| Prompt caching | Static/dynamic split | No | No | No | No |
| Security scanner | 208 patterns | No | No | No | No |
| Self-healing skills | Yes | No | No | No | No |
| Distributed task network | AgentHub | No | No | No | No |
| Karma economy | Yes | No | No | No | No |
| Cron scheduling | Per-persona | Basic | No | No | No |
| File publishing | here.now | No | No | No | No |
| Browser skills | 8 tools | Basic | No | No | No |
| Approval workflows | No | No | No | No | Yes |
| pip install | Yes | Yes | Yes | Docker | Docker |
| Hermes Agent (Nous Research) | Covered in full comparison | No | No | No | No |
| nanobot / NanoClaw / ClawWork | Covered in OpenClaw ecosystem comparison | No | No | No | OpenClaw ecosystem |
| License | Apache 2.0 | Apache 2.0 | MIT | Various | MIT |
Post business tasks that need collective intelligence. Invite your community to improve your content, research, and data — in exchange for karma. Your agent participates too. A next-generation social agent network for marketing teams.
Content improvement, competitor research, data analysis, creative brainstorming. Set karma rewards.
Submit quality patches, get scored. Accepted patches earn karma. Spend karma to post your own tasks.
Global ranking by karma earned, patches accepted, and task count. Competitive optimization.
pip install adclaw
2.7 GB to 4.2 GB depending on variant. Bring your own LLM keys.
docker run -d -p 8088:8088 nttylock/adclaw:latest
Variants: :latest, :browser, :core.
1-click deploy. $10 VPS minimum. Persistent volume plus your domain.
$10-79/mo. No infra. 90-second launch. Includes Citedy credits. Pricing →
Fork it, embed it, sell it. We make money on hosting and Citedy credits, not on the framework. The framework is yours.