Our take
Newless than six months old and changing fast- Activity
- 79
- Community
- 76
- Issues
- 68
- Pull requests
- 82
+4,038 stars in the last 29 days
- Merges more pull requests than 92% of the projects we track
OpenBot is a promising, actively developed alpha for teams willing to self-host and customize an agent platform, with AG-UI interoperability and action governance as its distinguishing strengths.
Good fit if
- You want to run AI coworkers on your own infrastructure and customize the repository as a template.
- Your agents can expose AG-UI endpoints, and you need policy checks and an audit trail around their computer and tool use.
- Your team can operate Docker, PostgreSQL, and CopilotKit Intelligence; the project is changing quickly and has contributions from a broad group without a single-contributor dependency.
Look elsewhere if
- You need a hosted product or a ready-made package rather than a repository to adapt.
- You need a mature, settled platform: OpenBot is explicitly alpha and its recent development pace indicates that behavior may change.
- You cannot use the required Docker services and external or self-hosted CopilotKit Intelligence setup.
All health signals
| Last commit | 2026-10-20 (today) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 19 (latest v0.2.0, 2026-10-12) |
| Contributors | 45 (top contributor: 27% of commits) |
| Issues closed, last 90 days | 5+ (typically closed in 1 day) |
| Pull requests merged, last 90 days | 89 (typically merged in 0 days) |
| Project age | 2 months |
Checked on 2026-10-21 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
10,171Forks
1,255Change since Nov 4, 2026: +129 stars·+19 forks
Overview
OpenBot is a self-hosted AI coworker platform built on AG-UI, CopilotKit’s open agent-to-user protocol. It lets teams bring agents built with different frameworks, or written by hand, into a shared web app without handing them ungoverned access to company tools. Each Bot gets its own computer, browser profile, files, and granted tools. A gateway evaluates and records actions before they are carried out.
The project is presented as a template to clone and customize, not a hosted service or a package to install. It runs on your infrastructure, with PostgreSQL for product data and CopilotKit Intelligence for durable threads and memory. It is aimed at organizations that want to adapt the interface, agent setup, and controls to their own environment.
Key Features
- Connect any AG-UI-speaking agent, independent of its framework.
- Give each Bot a separate browser, workspace, and browser profile.
- Apply fail-closed CEL policies to browser, file, shell, MCP, and component actions.
- Record permitted, refused, and failed actions in an audit trail.
- Let a person take control when a Bot reaches a login or two-factor prompt.
- Govern MCP tools, credentials, and reusable components through the app’s admin surfaces.
- Create scheduled routines and connect coworkers through channels.
- Run locally with Docker Compose, or deploy using a Docker image or Helm chart.
Use Cases
- A company wants an internal assistant that can browse its tools but must ask or be blocked before sensitive actions.
- A team has agents built with different frameworks and wants to connect them through one interface and governance layer.
- IT or security administrators need to inspect what agents attempted, including refused actions and the rules behind them.
- Teams want to assign recurring tasks to an agent while retaining per-action controls and an audit record.
What you need
Detected in the repository
- Node.js (from package.json)
- A Dockerfile, so it can run in a container; a Compose file sets up the related services
- A Helm chart for Kubernetes
- A test suite and automated checks on GitHub Actions
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
cp .env.example .env
bun install
Configure the Intelligence project and model key, then start the stack:
bash scripts/start.sh
Open http://localhost:3010. See the README for setup details and configuration, and the documentation for deployment guidance.
Alternatives
- omnigent: Omnigent orchestrates coding-agent runtimes with policies and shared sessions, while OpenBot provides AG-UI coworker agents with isolated browsers and files.
- BrowserOS: BrowserOS focuses on agents using local logged-in browser tabs, while OpenBot provides a self-hosted agent workspace with per-agent files and governed actions.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| OpenBot | TypeScript | MIT | 6.8k | New |
| omnigent | Python | Apache-2.0 | 11k | New |
| BrowserOS | TypeScript | AGPL-3.0 | 14k | Active |
Considerations
OpenBot is new and marked alpha, so teams should expect rough edges and ongoing changes. Recent development is active, with frequent commits and releases, contributions from dozens of people, and no measured contributor concentration; however, only a small number of issues were closed in the measured period. Setup also involves more than the app: Docker is required, and the project depends on CopilotKit Intelligence for durable threads and memory. No model ships with OpenBot, so administrators must configure model credentials. A fresh local setup defaults to single-user administrator access; deployments intended for others need sign-in configured. The repository is MIT-licensed.
Found this useful?
Share it with someone who would like OpenBot.