Outerlimit Raises $16 Million to Secure AI Agents at Execution

Outerlimit raised $16 million in pre-seed funding as the company emerged from stealth with technology designed to secure AI agents at the point of execution, where software can interact with tools, systems and data.

The round, announced Sept. 22, was backed by AlbionVC, Evolution Equity Partners and Crane Venture Partners. Outerlimit is developing a decentralized security and authorization layer for AI agents that perform multi-step tasks with limited human intervention.

The platform is organized around three functions. It discovers agents, tools, Model Context Protocol servers and shadow AI; observes agent behavior while preserving multi-hop chain integrity; and enforces controls when agents attempt to take actions.

Its architecture fragments credentials, keys and secrets across the agent ecosystem. Those elements are reconstructed only after verifying an agent’s identity, policy and execution context.

“Blocking adoption isn't a strategy, it simply drives the use of unsanctioned AI outside of enterprise oversight,” CEO Tony Pepper said.

Pepper founded Outerlimit with Neil Larkins and CTO Dr. Peter Vincent. Pepper and Larkins previously led Egress Software, which KnowBe4 acquired in 2024.

Vincent said agents can change their behavior based on information they read or interactions with other agents while operating at machine speed. The company’s approach binds identity, authorization and action during execution.

Outerlimit works with Fortune 500 and FTSE 100 global brands, but the company has not identified those customers or provided deployment metrics. Strategic angel investors include Charles Gorintin, Sam Morgan, Brian Murphy, Scott Price, Kyle Griswold, Nicola Sinclair, David Garfield and Manish Madhvani.

The company operates in London and New York. Outerlimit has not disclosed production-performance data or figures on incidents prevented.