
AI agents can now book travel, write code and spend money on your behalf. That power is also the problem. A new startup called Runta has raised $20 million to keep those agents from doing damage.
Andreessen Horowitz (a16z) led the round, which values Runta at more than $100 million, The Information reported.
‘Parenting’ AI agents
Runta’s founder, Guanlan Dai, is a father of two, and he sees AI agents a little like precocious children. Parents childproof the house and keep the credit card out of reach. Developers, he argues, must limit which files an agent can touch and how much it can spend at once.
His company helps businesses “parent” their agents. It wraps them in isolated environments and guardrails, so a rogue agent cannot run up a bill or wreck a system.
A computer for every agent
Runta is trying to rebuild the layer where agents actually run. In a16z’s post announcing the deal, partner Martin Casado said agents “just want a computer.” He means a full operating system, stateful, able to run locally or in the cloud, with security controls built in.
Casado argues this is not “yet another sandbox cloud” but core systems software rebuilt for agents. He also flags an odd side effect of the agent boom. A CPU shortage now sits alongside the model-driven GPU crunch, because agents need a lot of ordinary compute.
Dai knows that layer well. He was a technical lead on Cloudflare’s edge team, then built the core proxy at API company Kong.
The enterprise agent land grab
Runta joins a growing scrum of startups selling the plumbing to run AI agents safely inside companies. Others are tackling agent authorisation, access governance and oversight of an AI workforce.
The pitch is the same across all of them. Businesses are handing real tasks to autonomous software. They want to keep it on a leash, and to limit the damage when an agent goes off the rails. Casado calls the move from hosting software to hosting agents the biggest shift in computing yet.
Published July 17, 2026 - 11:43 am UTC
Back to top
View original source — The Next Web ↗

