Repolode
Submit
moltis-org

moltis-org/moltis

A self-hosted personal AI agent server in a single Rust binary, with sandboxed tool execution, multiple LLM providers, and chat channels like Telegram and Discord.

Open-source alternative to:ChatGPT

View on GitHubWebsite
Stars
2,908
Forks
359
Open issues
119
Last commit
4 hours ago

What is it?

Moltis is a persistent personal agent server that runs as one Rust binary on your own hardware, such as a Mac Mini, a Raspberry Pi, or a server you control. It connects to multiple LLM providers, keeps session state and memory in local storage (JSONL sessions, SQLite with full-text and vector search), and can be reached through a web UI, Telegram, Signal, Discord, Slack, Microsoft Teams, and other channels.

Tool execution runs inside Docker, Podman, Apple Container, or WASM sandboxes rather than on the host. Keys are stored in an encryption-at-rest vault, and authentication supports password, passkey, and API keys. The project also covers voice input and output, MCP servers over stdio and HTTP/SSE, a skill system, scheduling, and a native desktop client.

The Rust workspace is split into 59 crates grouped by role, so each part can be read or audited separately. A lightweight feature set is available for constrained devices.

Key features

  • 01Single Rust binary with no Node.js or npm runtime required
  • 02Tool execution in Docker, Podman, Apple Container, or WASM sandboxes
  • 03Multi-provider LLM support with streaming and sub-agent delegation
  • 04Chat access via web UI, Telegram, Signal, Discord, Slack, and Teams
  • 05Hybrid vector and full-text memory stored in SQLite
  • 06MCP server support over stdio and HTTP/SSE

Use cases

  • Running a private AI assistant on a home server or Raspberry Pi
  • Chatting with an LLM agent from Telegram or Discord while keeping data local
  • Giving an agent controlled, sandboxed access to shell commands and tools
  • Deploying a self-hosted agent server to Fly.io or DigitalOcean

Topics

#ai-agent#self-hosted#rust#llm#mcp#sandbox#telegram-bot#voice-assistant#personal-agent#ai-assistant

FAQ

What is moltis?
A self-hosted personal AI agent server in a single Rust binary, with sandboxed tool execution, multiple LLM providers, and chat channels like Telegram and Discord. Moltis is a persistent personal agent server that runs as one Rust binary on your own hardware, such as a Mac Mini, a Raspberry Pi, or a server you control. It connects to multiple LLM providers, keeps session state and memory in local storage (JSONL sessions, SQLite with full-text and vector search), and can be reached through a web UI, Telegram, Signal, Discord, Slack, Microsoft Teams, and other channels.
Is moltis free to use? What license does it have?
moltis is open source under the MIT license. Check the license terms before using it in commercial products.
Is moltis still maintained?
Actively maintained. The last commit to moltis was on Oct 22, 2026.
What can I use moltis for?
Running a private AI assistant on a home server or Raspberry Pi Chatting with an LLM agent from Telegram or Discord while keeping data local Giving an agent controlled, sandboxed access to shell commands and tools Deploying a self-hosted agent server to Fly.io or DigitalOcean
Where can I get moltis?
The source code and installation instructions are on GitHub: https://github.com/moltis-org/moltis

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects