Self-hosted AI workspace
Conversation and app state use local SQLite storage. Provider credentials are encrypted at rest.
Open source · self hosted · MIT licensed
An open-source personal AI agent workspace you can self-host
Looking for a self-hosted AI agent or an open-source alternative to OpenAI Dots, Meta Muse, Grok Bot, Instinct, Manus Cue, Claude Cowork, or ChatGPT agent? Open Dots is an MIT-licensed workspace for chat, selected tools and connectors, computer tasks, and approval-gated actions. Run it yourself, inspect the code, and adapt it to your workflow.
Early prototype for local experimentation. Review the documented security and deployment limitations before hosting it for others.
Visible action controls
A deny-by-default action gateway routes higher-risk operations through approval prompts and audit events.
Open implementation
MIT-licensed code, a configurable inference adapter, Composio connectors, and an optional Docker/Playwright computer runtime.
Compare Open Dots with personal AI agents
Looking for an open-source alternative to one of these products? Read the product-specific comparison, including what Open Dots does and does not currently replace.
- Open Dots vs OpenAI DotsOpen-source alternative overview →
- Open Dots vs Meta MuseOpen-source alternative overview →
- Open Dots vs Grok BotOpen-source alternative overview →
- Open Dots vs InstinctOpen-source alternative overview →
- Open Dots vs Manus CueOpen-source alternative overview →
- Open Dots vs OpenClawOpen-source alternative overview →
- Open Dots vs Claude CoworkOpen-source alternative overview →
- Open Dots vs ChatGPT agentOpen-source alternative overview →
What Open Dots can do today
Create assistant personas, stream chats, attach images, select configured models, and connect supported apps through narrowly scoped actions.
Request workspace and computer operations through an action gateway. Use the optional computer provider in Docker or connect a compatible remote service.
Open Dots is independently built and is not affiliated with or endorsed by OpenAI, Meta, xAI, Instinct, or Manus. It does not currently include persistent long-term memory, scheduled routines, a mobile app, multi-user roles, or a hardened sandbox for arbitrary web content.
Is Open Dots the right AI agent alternative for you?
Choose Open Dots if you are comfortable running software yourself and want to inspect an AI workspace with visible approval steps. It is not the best fit if you need a polished mobile assistant, persistent memory, scheduled work, broad everyday app access, or a managed computer that keeps running in the cloud.
Questions about this open-source AI agent
Can I self-host Open Dots?
Yes. The API and web client can be run on infrastructure you control. You’ll need to configure a compatible inference service and follow the setup and security notes in the README.
Which AI models does Open Dots support?
The app lets you configure model IDs, but the bundled inference adapter expects a specific prediction API contract. It is not a universal connector for every model provider or OpenAI-compatible API.
Is Open Dots a replacement for these personal agents?
It is an open-source project to evaluate when self-hosting and inspecting the implementation matter. It is still a prototype and does not match the feature set or convenience of the managed products.
Is Open Dots safe for unattended computer use?
The project routes higher-risk operations through approvals, but its optional computer runtime is not hardened for hostile websites. It is intended for local experimentation; review the limitations before deployment.