Clawlet
Self-hosted AI agent with local memory search, Ollama and vLLM support, and chat app integrations. Runs as a standalone binary or in Docker under MIT.
Clawlet is a self-hosted personal AI agent for people who want an assistant on their own machine or server, with searchable memory across conversations. Written in Go, it ships as a standalone binary with no separate runtime dependency. It supports macOS on Apple Silicon and Docker deployment, and it's open source under the MIT license.
You can use a local LLM through Ollama, vLLM, or another OpenAI-compatible endpoint. Cloud choices include OpenAI, Anthropic, Gemini and OpenRouter, plus OpenAI Codex through OAuth login. The agent runs on your hardware; choosing a cloud model sends model requests to that provider. Memory embeddings can also use a local endpoint, such as Ollama with nomic-embed-text, or OpenAI's embedding service.
Its memory system combines semantic search with local storage using bundled SQLite and sqlite-vec. Optional search tools let the agent retrieve past context from Markdown memory files, while the search index stays in the workspace. You can chat in the terminal or through Telegram, WhatsApp, Discord and Slack. Scheduled tasks can run once or recur and deliver messages to a connected chat.
Clawlet accepts image, audio and text attachments. It passes images to vision-capable models and uses the selected provider for audio transcription. File access stays within the workspace by default, and the gateway listens only on localhost unless public access is explicitly enabled. Shell execution includes guards against unsafe commands and access to sensitive paths.
