🦞 Headless real-time Python voice assistant powered by ElevenLabs, Gemini, Mem0, MCP tools, and local wake-word detection.
A lightweight, open-source, voice-first personal AI agent.
HeyClaw listens locally for a wake word, opens a real-time voice session, and connects speech to a small, readable Python agent built around Gemini, Mem0, MCP tools, and workspace skills. It keeps the device-facing audio path separate from the agent backend, so the intelligence can run remotely while each room only needs a lightweight satellite.
- Voice first: wake word, microphone input, streamed responses, and interruption-aware speech are the primary interface. - Small, readable core: the agent loop is intentionally compact and avoids a large orchestration framework. - Persistent memory: Mem0 retrieves relevant user context and stores durable details across conversations. - Tool ready: MCP servers provide external capabilities, while workspace skills define when and how the agent may use them. - Self-hosted orchestration: the backend, prompts, workspace, tools, and configuration remain under your control. - Satellite architecture: audio capture and wake-word detection are isolated from the backend and ready to move onto dedicated hardware.
- local wake-word detection with openWakeWord; - real-time speech-to-text and text-to-speech through ElevenLabs Speech Engine; - low-latency responses with Gemini 3.1 Flash-Lite; - a DSPy ReAct agent with dynamically discovered MCP tools; - long-term, user-scoped semantic memory through Mem0; - runtime instructions, identity, user profile, and skills loaded from a local workspace; - a FastAPI service for health and readiness checks; - separate backend and satellite Python pa
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/vericle--heyclaw.html)
Self-hosted, local-first AI assistant server: swappable LLM providers, review-gated memory, an encrypted vault, and wake-word voice. No accounts, no telemetry.
Alex663028/pulse-agentPulse — Hermes-style self-improving AI agent. Reliability-first rebuild with evaluated skill self-evolution, multi-agent team orchestration, dialectic user modeling, and fully self-hosted default stac
charbelkassab/relatio-crmLightweight self-hosted CRM with a built-in MCP connector. Let Claude, ChatGPT, or Gemini run your pipeline.
bybit-exchange/kaasTurn scattered notes, docs and transcripts into a queryable Markdown wiki — an LLM knowledge-base compiler with MCP access, no embeddings, self-hosted.
zja0011/edgesentinel-visionopsJetson Nano edge AI vision Agent Harness: real-time video analytics, MCP tools, event evidence, DeepSeek online/offline, secure recovery
rokbenko/quackd🦆🧠 Give your Microduck a brain. Tell a small robot with two legs what you want in plain language. An LLM (Claude, OpenAI, Gemini, Grok) uses the skills it already has to do it. Includes a simulator, .
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.