a16z is spearheading a $20M seed round for the AI agent startup Runta.

      AI agents are now capable of booking travel, writing code, and spending money on your behalf. While this capability is powerful, it also presents risks. A new startup named Runta has secured $20 million in funding to mitigate the potential threats posed by these agents. According to The Information, the investment, led by Andreessen Horowitz (a16z), has valued Runta at over $100 million.

      "Parenting" AI agents

      Runta's founder, Guanlan Dai, who is a father of two, likens AI agents to precocious children. He believes that, much like parents childproof their homes and keep credit cards out of reach, developers should restrict the files an agent can access and limit its spending capacity. His company assists businesses in "parenting" their agents by enclosing them in isolated environments with guardrails, preventing rogue agents from incurring excessive charges or damaging systems.

      A computer for each agent

      Runta is focused on redesigning the fundamental layer on which agents operate. In a16z's announcement of the deal, partner Martin Casado stated that agents "just want a computer." This refers to a complete operating system that is stateful, capable of functioning locally or in the cloud, and integrated with security controls. Casado emphasizes that this is not simply "another sandbox cloud" but rather core systems software specifically rebuilt for agents. He also points out a notable consequence of the rise in agents: a CPU shortage now accompanies the ongoing GPU crunch, as these agents require substantial standard computing power.

      Dai is familiar with this layer, having served as a technical lead on Cloudflare’s edge team before developing the core proxy at the API company Kong.

      The enterprise agent land grab

      Runta is entering a rapidly expanding field of startups providing the infrastructure necessary to manage AI agents securely within organizations. Other companies are addressing issues related to agent authorization, access governance, and oversight of AI workforces. The common theme among these ventures is the recognition that businesses are entrusting genuine responsibilities to autonomous software. They aim to maintain control and minimize potential damage when an agent acts unpredictably. Casado describes this transition from hosting software to hosting agents as the most significant evolution in computing thus far.

      Published July 17, 2026 - 11:43 am UTC