Anil-matcha/open-dots
Open-source, self-hosted AI agent workspace and alternative to OpenAI Dots, Meta Muse, Grok Bot, Instinct, Manus Cue, Claude Cowork, and ChatGPT agent. MIT-licensed; includes chat, connectors, approvals, and optional computer use. Early prototype.
What it solves
Open Dots provides a self-hosted, open-source workspace for managing personal AI agents. It addresses the need for a local-first alternative to proprietary agent platforms (like OpenAI Dots or Meta Muse) by giving users full control over their conversation history, model selection, and the execution of high-risk actions through a governed action gateway.
How it works
The project consists of a Next.js frontend and a FastAPI backend. It uses a configurable inference adapter to connect to various LLM providers and a SQLite database for local state. To ensure security, it implements a "deny-by-default" action gateway that requires explicit user approval for higher-risk tasks. For computer automation, it offers an optional Docker/Playwright runtime that creates isolated workspaces per assistant.
Who it’s for
Developers and individuals who want an inspectable, self-hosted AI agent environment with local data storage and explicit control over what the AI can do on their computer.
Highlights
- Governed Action Gateway: Prevents unauthorized AI actions by pausing for approval and creating audit events.
- Customizable Personas: Create assistants with unique instructions, model IDs, and visual identities.
- ** لقب Local-First Storage**: Persists conversations and application state locally using SQLite.
- Computer Runtime: Optional support for Docker-based browser automation via Playwright.
- Web Search Integration: Built-in
/searchcommand using the You.com MCP server for real-time information. - Third-Party Connectors: Integration with Composio for specific app connections like GitHub issue management.
Related
- Project
- Project
- Project
- Project