other / agent

OpenBot

A template for running AI coworkers on your own infrastructure, each with its own browser and files, plus a policy gateway and audit log for its actions.

At a glance

Purpose
A template for running AI coworkers on your own infrastructure, each with its own browser and files, plus a policy gateway and audit log for its actions.
Runs in
Local agent with shell, Web service
Operating systems
macOS
Needs
Runtime: Docker and Bun 1.3+; Needs internet: Yes; Needs accounts or API keys beyond your agent: Yes
Files and data
Reads files: Yes; Writes files: Yes; Sends data to other services: Yes
Tool type
Agent
Delivery
From GitHub
Version
Latest on the main branch
Access
Code on GitHub
Compatibility
According to the README, OpenBot runs with Docker and Bun, and ships as a single container image as well as a Docker Compose setup. The README says any AG-UI endpoint can serve as a Bot and names LangGraph, Mastra, CrewAI, Pydantic AI and Google ADK. Model keys it names are OpenAI, Anthropic and Google, with base URL settings for gateways or proxies. Sign-in providers it names are Google, Microsoft and Okta, plus company SAML or OpenID Connect providers. The README describes continuing Bot conversations in Slack, Microsoft Teams or by text message. The README names Google Drive, Notion and Parallel Search in the MCP catalogue and says Composio brokers additional apps. The README says Bot computers can run under gVisor where the host supports it.
External costs
The README requires a CopilotKit Intelligence project and licence and a model key. The README names a free Intelligence plan, paid Intelligence plans, and a free Developer plan for local Intelligence evaluation on a Mac. The README offers CopilotKit engineering services to install and customize OpenBot for a company. The README gives no prices for those services or plans. Your coding agent or model provider may charge for usage.
Permissions
The README says each Bot has a container, Chromium profile, workspace and shell, and can use real browser logins. The README says browser/file/shell/MCP actions pass an audited policy gateway; approval settings determine human gates. The README describes takeover and admin-stored credentials. The README shows bun dependencies and a helper provisioning an Intelligence runtime key and Learning container and writing .env. The README says startup launches Docker services, applies migrations and starts the API, routine worker and app. The README says the supervisor holds the Docker socket. The README describes scheduled routines acting as the requester and conversations through Slack, Microsoft Teams or texts. The README says managed Intelligence is the quick-start route; self-hosting uses INTELLIGENCE_API_URL and a matching gateway setting.
Usage terms
MIT License, as stated in the repository. See its LICENSE file for the exact terms.
Interface
web

Code on GitHub

CopilotKit/OpenBot

MIT

TypeScript6,409 stars875 forksLast pushed 13 Oct 2026 UTC

Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.

Topics: ag-ui agent-governance ai-agents browser-automation copilotkit generative-ui mcp

Public GitHub data fetched · 0 days old

Skill folders in this repository (2)

Folders that contain a SKILL.md file, from GitHub's file list on .

View on GitHub ↗

What it does

OpenBot is a template for self-hosted AI coworkers called Bots. The README labels it alpha and under active development, and says it is intended to be cloned and customized. The README says Docker Compose starts its services, PostgreSQL holds product data, and an administrator supplies a model credential.

The README says each Bot has a container with Chromium, a browser profile and a workspace volume. The README describes screen viewing and human takeover, and a gateway that evaluates CEL policy rules and records browser, file, shell and MCP actions before dispatch.

The README says AG-UI endpoints can serve as Bots and names LangGraph, Mastra, CrewAI, Pydantic AI and Google ADK. The README names General Assistant and Knowledge as starting coworkers, and lists channels, approvals, skills, routines and administration pages.

Who it is for

Teams and developers who want to host AI coworkers on their own infrastructure and are willing to clone and customize a template.

Example tasks

Ask a Bot to open a website in its browser and report what it finds.
Open the audit page to see permitted, refused and failed actions.
Add a deny rule or preset on the boundaries page and retry the same browser action.
Create a Bot from the Bots page, give it a standing role and start a channel with it.
Register your own AG-UI endpoint as a Bot.

Getting started

According to the README, copy the environment template and install dependencies with cp .env.example .env and bun install. Connect a CopilotKit Intelligence project with the login, project select and setup commands the README gives, fill in OPENAI_API_KEY in the env file, then run bash scripts/start.sh and open the app on local port 3010. The README says the start script starts the Docker services, applies migrations, starts the API server, the routine worker and the app, and checks their health routes. It also describes running a published container image for deployment. The README has the exact commands.

Limitations

The README labels OpenBot alpha, under active development and a template to customize. The README requires Docker, Bun 1.3+, a CopilotKit Intelligence project and licence, and a model key. The README says the proof-of-concept Bot uses OpenAI and the LangGraph Bot can use OpenAI, Anthropic or Google. The README says fresh self-hosted Intelligence setup opens an isolated Chrome or Edge window and requires either browser installed. The README describes local Intelligence as a macOS Docker Desktop preview rather than a production installation, with a 30-day local licence renewable while the account is in good standing. The README says the environment template admits requests as one administrator; sign-in must be configured before other people can reach the deployment. The README tells operators to put TLS in front of remote deployments and keep lower-level computer services private. The repository can change at any time; check its README before you install.

Data practices

The README says PostgreSQL stores product data, policy, audit records, credentials and grants, while CopilotKit Intelligence stores durable conversation threads and memory. The README's quick start uses managed Intelligence URLs by default; the README describes changing the API and gateway settings for self-hosted Intelligence. The README says Bot files and browser profiles persist in volumes. The README says stored credentials are encrypted at rest and redacted from audit events. The README says the secret-request trail records that a secret was requested and its length. The README describes model credentials and connected applications; data sent to those services depends on the configured connections.

Community experience

Not rated yet

Sign in to share your experience.

Comments 0

No comments yet.

Sign in to comment

Report a listing concern