MantisClaw
Autonomous agents controlled from WhatsApp.
Run, control and supervise AI-driven operations directly from WhatsApp conversations. Direct messages and group chats become a command interface for real task execution.
MantisClaw turns WhatsApp into the control surface of a fully capable autonomous agent. Send a command, start a task, receive the result — a Word document, an Excel sheet, a report, a screenshot, a completed browser workflow — back in the chat.
It is local-first: operational data lives in a portable PostgreSQL, secrets are injected at runtime and never shared with the model, and external AI providers are optional. Windows, Linux and macOS builds ship from the same pipeline.
- Multi-user, multi-account, multi-channel: flexible chat → agent mappings
- Python kernel, browser agent (Playwright), file output (Excel / Word / PPT / PDF)
- Meeting recorder for Google Meet, Teams and Zoom
- Local PostgreSQL, no forced cloud dependency
- Windows, Linux, macOS (Intel and Apple Silicon)
WhatsApp command surface
Direct and group chats. Teams collaborate while supervising agent operations.
Real execution
Writes and debugs code, runs scripts, calls APIs, drives a browser, produces documents.
Secure runtime
API keys, credentials and tokens are injected at runtime, never exposed to the LLM.
Skills & scheduling
Editable skill system, playbooks and autonomous scheduled tasks.
Architecture: WhatsApp → router → agent core → execution layer
Meeting recorder
Agent dashboard
Multi-platform build pipelines
- Channel
- WhatsApp direct and group chats
- Runtime
- Go API server, embedded Python, Playwright
- Data
- Portable PostgreSQL, local
- Models
- OpenAI-compatible or local (Ollama)
- Platforms
- Windows, Linux, macOS
Ready to put AI on your own infrastructure?
Tell us about your environment, your data boundary and your latency target. We will design the pipeline with you and bring it live within days, not months.