The Enterprise Agentic Frontend Stack. Build rich agentic applications that learn from every interaction. Self-hostable and built to scale with your org.
This is the 5th launch from CopilotKit. View more
OpenBot
Launching today
OpenBot is the modular, self-hostable, enterprise-ready agent workspace with a clean architecture and built-in governance. Use it as-is, white-label it, or deeply customize it. Bring the agents your engineers build or start with ours, then choose the models, tools, and infrastructure they run on.
Free Options
Launch Team
AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted
Hey Product Hunt 👋
I work with the CopilotKit team, and I’m excited to share OpenBot.
Teams are experimenting with AI agents, but putting them to work internally raises harder questions: which company data can each person access, what can an agent do in a browser, and where does the work run?
OpenBot is an open-source, self-hosted workspace for employees and custom agents. You bring the models and agents built on your stack. Any agent that speaks AG-UI can connect as an endpoint, with its own channel.
What you can do:
• Give each agent a persistent thread, so the conversation and its work stay together.
• Connect Google Drive or OneDrive so agents can answer from company documents with sources attached. Retrieval follows the requesting employee’s permissions. If a permission mapping is unclear, the document is withheld.
• Let agents use a browser for systems without a useful API. Watch the screen, take control mid-task, then return to the same conversation. Each agent can have its own computer, files, and signed-in browser.
• Gate sensitive MCP calls with per-tool allowlists and keep a record of each call and outcome.
• Run OpenBot on infrastructure your organization owns. Documents, vectors, permissions, and conversations live in your Postgres. Admins hold the provider credentials. Docker Compose gets a demo running; Helm is the supported path to production.
OpenBot is built for teams that want useful internal agents with clear boundaries around data, tools, and deployment. Because it’s built on AG-UI, you can bring agents from LangGraph, Mastra, Pydantic AI, or your own stack.
Try it at https://open.bot.
If your organization is evaluating agents, which boundary is hardest to solve: data permissions, browser access, or deployment and credential ownership? What controls would you need before rolling an agent out to employees? We’d love to hear your perspective.