Open Dots is an open-source, self-hosted workspace for running AI agent tasks — an alternative to paid personal-agent products such as OpenAI Dots, Meta Muse, Grok Bot and Claude Cowork. It wraps the Claude Code CLI in a workspace you control, with a web interface, a Telegram interface and a desktop app. The project has about 5,900 stars on GitHub and is in active development in 2026.
You give Open Dots a task from Telegram or the web UI; it runs the coding agent inside a disposable sandbox using your own locally signed-in Claude credentials, and shows you risky actions for approval before they execute. It can connect to GitHub, run prompts on a schedule, and keep named agents for recurring kinds of work. The author is candid about its stage: it currently supports Claude only, has no multi-agent features yet, and is an early-stage project rather than a feature-for-feature replacement for the commercial products it takes inspiration from — the roadmap and known-limitations list are published in the repository.
The project’s emphasis on approvals and disposable sandboxes reflects the trust problem facing personal agents generally: people want an agent that acts, but not one that acts unsupervised on their real accounts and files. Open Dots’ answer is to keep everything local and gated — the agent proposes, you approve, and each task runs in a sandbox that can be thrown away.
Free. Open Dots is open source; you bring your own Claude subscription or API access, which is what actually powers the agent.
Developers who like the personal-agent concept of Dots or Cowork but want to host the workspace themselves and drive it from chat, with approvals in the loop.