OpenBot

OpenBot

Nightly Labs · Coding · Productivity · Chatbot

OpenBot is a free, open source desktop workspace that runs Codex, Claude, Gemini and Grok as persistent AI teammates on your own computer, using the AI plans you already pay for.

Interface preview of OpenBot

About OpenBot

What Is OpenBot

OpenBot is a desktop app that turns the AI subscriptions you already hold into a small team of agents working on your machine. Each agent gets a name, a set of instructions, and a folder of its own. It reads and writes files there, runs commands, and opens pages in a browser built into the app.

The point is continuity. Close the window and the agent keeps its conversation and its workspace, so you don't explain the project again every morning. Agents can also message each other, hand off a task, and pass files along. That's why the app is built around a team rather than a single chat box.

The app costs nothing. You pay your AI provider through the plan or API key you already have, or you run a local model through Ollama or LM Studio. The source code sits on GitHub under a noncommercial license, and the project is still labeled a development preview.

Getting Started

  1. Download the build for macOS, Windows or Linux from the official site.
  2. Choose a provider and sign in once. OpenBot downloads and pins the command line tool it needs, or you point it at an OpenAI compatible endpoint.
  3. Describe the agent you want in one prompt, read the instructions it drafts, then save it with a name.
  4. Send the first task. Anything you send while the agent is busy waits in a queue you can pause or cancel.

Product Information

A quick look at OpenBot's pricing, supported platforms, and performance.

Free PlanYes
Paid Plans$0 app; hosted server from $25/mo
PlatformmacOS 13+, Windows 10+ (x64), Linux (x64 and arm64 AppImage), iOS, Android, web
DeveloperNightly Labs
CategoryCoding · Productivity · Chatbot
Release DateAug 2026
Latest UpdatedNov 2026
Website Visits5.3K
Website Global Rank4.4M
API AvailabilityN/A

Best for

The users, tasks, and scenarios where this tool fits best.

Users

  • Developers who already pay for several AI tools
  • Small teams without an internal platform
  • Freelancers with client material on their laptop
  • Students and hobbyists on a tight budget

Tasks

  • Multi step coding work
  • Research that needs a browser
  • Recurring admin chores
  • Moving a long job between providers

Scenarios

  • Solo work on a laptop
  • A desktop that stays on at home or in the office
  • Handing off a stuck session

Key features

Agents that work as a team

Every agent has its own name, instructions and workspace, and can send messages to the others. One agent can hand a finished draft to a reviewer and wait for the reply. Files move through a shared area, so you don't paste the same document into three chats.

Use the AI plan you already pay for

Codex signs in with a ChatGPT plan, Claude Code with a Claude plan, Gemini with Google AI Pro or Ultra, and Grok with an account or an xAI key. OpenCode, Cursor, Cline, Muse Code, Pi and Inception's Mercury models are also on the list, along with any agent that speaks the Agent Client Protocol. Each agent can run a different model, which makes it easy to put a cheap model on bulk work.

Files and conversations stay on your computer

Workspaces, chats, attachments and browser data live on the machine that runs OpenBot, not on the vendor's servers. Your provider still receives the prompts an agent sends, and pages opened in the browser still use the network. For anyone handling client code or unpublished drafts, that split is the part that matters.

Change the provider, keep the agent

An agent keeps its workspace and conversation when you restart the app or move it to another provider. If a plan reaches its limit, or a provider is overloaded, the agent can switch to a backup model and shows a note in the chat. You don't rebuild the setup or re-upload the files.

A browser and computer control built in

Agents open, read and control pages inside a browser that comes with the app. An optional Computer Use mode goes further and lets an agent move the cursor, click and type in other desktop apps. On macOS that mode needs screen recording and accessibility permissions, which you grant once per machine.

A queue for work that arrives mid task

You can keep sending instructions while an agent is busy. Messages wait in a queue you can pause, resume or cancel, and the queue survives a restart. That turns a single agent into something closer to a colleague with a backlog, instead of a chat window that refuses input.

Pros and cons

Pros

  • The app is free and has no locked features, so the only recurring cost is the AI plan or API key you choose.
  • Provider choice is unusually wide, from ChatGPT and Claude subscriptions to local models in Ollama.
  • Data stays on your own disk, which suits code, contracts and anything under an NDA.
  • Agents keep their workspace and history across restarts and provider switches.
  • The source code is public, so a team can audit what the agents actually run.

Cons

  • The project is a development preview, so expect breaking changes and occasional rough edges in a tool you may depend on daily.
  • Agents run with broad permissions and don't ask before each file change or command, which puts the burden of scoping tasks on you.
  • Local first doesn't mean offline. Without a local model, an agent stops working when the network drops.
  • The license covers noncommercial use only, so a company that wants to use it commercially needs a separate agreement or the hosted server.

Frequently asked questions

Yes. The app costs $0 and no feature sits behind a paywall. You still pay your AI provider through the subscription or API key you connect, and a hosted team server starts at $25 a month.