OpenAI CEO Sam Altman said on February 16 that Peter Steinberger, the founder of open-source AI agent project OpenClaw, is joining OpenAI to work on next-generation personal agents. Altman called Steinberger a “genius” and said multi-agent collaboration will soon become a core part of OpenAI’s products. At the same time, OpenClaw will move into foundation-based stewardship as an open-source project, with continued support from OpenAI.
Steinberger moves from independent builder to OpenAI
Steinberger is an Austrian software engineer who founded PDF software company PSPDFKit in 2011. In 2021, the company was sold to Insight Partners at a €100 million valuation. After that exit, he launched Amantus Machina and started building a personal AI agent project. The project was first called Clawdbot, then renamed Moltbot after objections from Anthropic over its similarity to Claude, before settling on the name OpenClaw.
Explaining his decision to join OpenAI, Steinberger said he wants to change the world, not build a large company. He described OpenAI as the fastest route to bring the technology to everyone. That makes his move less like an acqui-hire and more like a direct bet on scale and distribution.
Why OpenClaw drew attention
OpenClaw is built to execute tasks rather than just chat. It runs on a user’s own infrastructure and uses messaging platforms such as Signal, Telegram, Discord, and WhatsApp as interfaces. From there, it can manage email, browse the web, book flights, handle calendar scheduling, and continue working while the user is asleep.
Its core feature is a “Heartbeat Scheduler,” which wakes the agent at configurable intervals so it can keep operating without constant human input. That design helped OpenClaw collect 180,000 GitHub stars at record speed, making it one of the most watched open-source AI projects in early 2026. The same architecture has also drawn scrutiny from security researchers because access to email, calendars, and messaging services creates privacy risk if the system is misconfigured.
OpenAI expands its multi-agent stack
Steinberger’s hire fits into a broader OpenAI push around AI agents. The company has recently introduced several related products: Agents SDK with support for the MCP protocol for external tool connections, ChatGPT Agent for autonomous computer use on complex tasks, the Codex app for managing multiple agents in parallel, and Frontier, an enterprise platform designed to let organizations manage AI agents much like employees.
Personal agents sit closer to consumer use cases than enterprise-focused Frontier or developer-facing Agents SDK. They are aimed at recurring daily work such as email handling, scheduling, and message management. OpenClaw’s open-source roots and already proven product shape give OpenAI a ready-made framework in that segment. Altman also said open source is an important component of the multi-agent future, a notable position given the company’s recent debates over open versus closed development.