Moltis

Moltis is a self-hosted personal AI agent server for people who want an assistant that remembers past conversations and can act on their behalf. It runs on your own hardware as a single Rust binary, with a browser interface and access through messaging apps. It's open source under the MIT license.

You can use Ollama, LM Studio or local GGUF models, including CPU and GPU execution and MLX on Apple Silicon. It also connects to Anthropic, OpenAI, Google Gemini and OpenAI-compatible services. The server and persistent memory run on your host; choosing a cloud model or cloud voice provider sends requests to that service. Moltis runs on macOS, Linux and Windows, supports Docker deployment, and can also run on a cloud host you control.

Its tools cover recurring tasks described in natural language, browser automation, and CalDAV calendar events and reminders. Memory combines semantic and full-text search for recall across sessions. Voice input and spoken replies work with local or cloud providers, while Telegram, WhatsApp, Discord, Slack, Matrix, Nostr and Microsoft Teams give you other ways to reach the agent.

Tool execution uses sandbox isolation by default, with permission controls for actions and filesystem access. Passwords, passkeys and scoped API keys protect access, and an encrypted vault stores credentials. MCP support connects additional tools; skills and hooks let you extend behavior. An OpenClaw import brings over conversations, skills and settings without modifying the original workspace.