openbot

OpenBot is a free desktop application that lets you run a team of AI agents locally on macOS, Windows, Linux, iOS, or Android. The app works with the AI subscription plans you already have, such as ChatGPT, Claude, Gemini, or Grok, or with any model you can reach through an OpenAI‑compatible endpoint or a local runner like Ollama.

The core idea is that each agent has its own name, instructions, and workspace, and the agents can exchange messages, hand off tasks, and share files within the same environment. You create an agent by describing it in a single prompt, confirm its instructions, and then add it to a workspace. Once the team is assembled, you can queue work for the agents; messages wait in a queue that you can pause, resume, or cancel. Agents keep their conversation history and workspace files even after you restart the app or switch providers, so the context is preserved across sessions.

OpenBot includes a built‑in browser that agents can open, read, and control, allowing them to interact with web pages as part of a task. The app stores all workspaces, conversations, attachments, and browser data on the computer that runs OpenBot, not on external servers. If an agent's primary model reaches its usage limit or becomes unavailable, the system can automatically switch the agent to a backup model while retaining its chat history and noting the change in the conversation.

The platform supports a wide range of models: Codex, Claude Code, Gemini, Grok, OpenCode, Cursor, Cline, Muse Code, Pi, and Inception's Mercury, plus any model that implements the Agent Client Protocol. You can also connect custom endpoints, local models via Ollama or LM Studio, and third‑party agents like Goose or Qwen Code. OpenBot's source code is publicly available on GitHub under a non‑commercial license, and the application itself is free to use for personal projects, study, or research. Commercial use requires a separate license.

Key features

• Persistent workspaces and conversations that survive restarts
• Queue system for assigning tasks to agents, with pause/resume controls
• Built‑in browser that agents can read and manipulate
• Ability to mix and match models from multiple providers in one team
• Automatic fallback to backup models when limits are reached
• Open source code available on GitHub for inspection and modification