OpenBot 1.0 offers an open-source, self-hostable AI coworker that eliminates vendor lock-in by letting companies control their own models, data, and governance policies.
AI-generated summary. Check the original reporting for context and accuracy.
This is the available feed excerpt, which may be shortened or in the publisher’s original language.
Most AI coworker products are vendor locked. They pick the model, host the data, and write the rules. CopilotKit is taking the opposite route. OpenBot 1.0, was released November 9th, an MIT-licensed, self-hostable agent workspace. What sets it apart is, a company brings their own models, agents and infrastructure. Governance is fully baked in, and every Bot action is checked against a policy before it runs. What is OpenBot? OpenBot is a modular agent workspace from CopilotKit, the team behind the AG-UI protocol. You give a Bot a task and watch it work. You review findings, interact with results and approve next steps in the same conversation. Each Bot gets a computer of its own. That is a container with its own Chromium browser, logins and files. It reaches only the tools you grant. Answers can come back as charts, checklists, tables and cards. In the new v1.0, Bots now reply in Markdown by default and draw components when they help. How does the governance gateway work? Every browser click, file write, MCP tool call and component goes through one gateway. It evaluates a CEL policy, writes the audit row first, then acts. Deny rules run before allow rules, and a broken rule refuses rather than opens. On top of that, the business tier adds: Single sign-on and SCIM sync for users and groups. The README documents Google, Microsoft (Entra) and Okta, plus SAML and OIDC. On-behalf-of tokens, usually scoped to a subset of the person’s permissions. Tool grants per Bot, per role or per person, with access requests to an admin. Policies that block actions or route sensitive work to a human for approval. Spend visibility per Bot, role, team or person, plus daily, weekly or monthly token limits. A full audit trail with SIEM export, multi-tenancy and an admin plane. EU AI Act content labeling, retention controls and deletion on request. CopilotKit itself holds SOC 2 Type II. Which agents and models can you bring? Any agent that speaks AG-UI plugs in. The site lists
An automated sorting aid based on source trust, keywords, and entities, on a scale of 1–12. High: 9–12; medium: 6–8; low: 1–5. It does not measure urgency, verify reporting, or indicate editorial endorsement.
Liquid AI’s open-source LiquidCodec enables efficient, real-time speech-to-text and text-to-speech by converting audio into predictable tokens for language models.
MIT Lincoln Laboratory’s three R&D 100 awards validate its significant, real-world technological innovations, reinforcing its reputation for impactful research and development.
LLMON, a new markup language for LLMs, is now integrated into IBM's Granite 4.1 models, improving how these AI systems process and generate structured data.