New York-based startup Outerlimit has emerged from stealth with $16 million in pre-seed funding to build what it calls a decentralized security and authorization layer for agentic AI — a system designed to discover autonomous agents running inside an enterprise, observe what they're doing, and stop them from taking harmful actions before those actions happen.
What Outerlimit Does
Outerlimit's platform is built around three functions: Discover, which finds every agent, tool, and MCP server operating in an environment — including "shadow AI" nobody signed off on; Observe, which provides visibility into agent behavior and preserves the integrity of multi-hop agent-to-agent chains; and Enforce, which applies deterministic, cryptographically-backed controls to every action an agent attempts to take.
Rather than trying to make agents behave through alignment, Outerlimit enforces boundaries externally. The company says its architecture fragments credentials across a distributed system and reconstructs them only when identity, policy, and context all check out — meaning a compromised or misbehaving agent can't simply reuse a broad, standing credential to do damage. As the company put it, if an agent is delegated a token, Outerlimit can guarantee its scope, its location, and the conditions under which it's used.
Co-founder and CTO Dr. Peter Vincent, a theoretical neuroscientist by training, framed the problem as one of architecture, not policy: agents change behavior based on what they read, and harnessing that requires a fundamentally new security model — not a patch on existing identity and access management tooling.
The Funding Round
The $16 million pre-seed round — reportedly one of the largest pre-seed raises on record in cybersecurity — was led by AlbionVC, Evolution Equity Partners, and Crane Venture Partners. Angel investors include Charles Gorintin (Alan co-founder and a Mistral AI advisor), Brian Murphy (ReliaQuest founder/CEO), Scott Price (A-LIGN founder/CEO), Sam Morgan (Goldman Sachs), Kyle Griswold (FTV Capital), and several others.
The Founding Team
Outerlimit was founded by CEO Tony Pepper and Neil Larkins, who previously built Egress Software Technologies and sold it to KnowBe4 (a Vista Equity Partners portfolio company) in 2024, alongside Dr. Peter Vincent as founder and CTO. Pepper has argued that security teams today "are being asked to sign off on risks they cannot adequately control" — a gap he says Outerlimit is built to close.
Why It Matters
Enterprises are handing autonomous agents real credentials and real permissions to act — booking transactions, querying internal systems, calling other agents — often faster than security teams can inventory what's running, let alone govern it. Traditional access-control models assume a human who understands consequences is behind every action; agents don't carry that same restraint, so identity and access management built for people doesn't map cleanly onto them.
Outerlimit's launch fits a broader 2026 wave of "agentic AI security" and "AI governance" startups racing to fill that gap, alongside vendors building agent identity, runtime monitoring, and policy-enforcement tools for the same problem. Whether a distributed, credential-fragmenting approach outcompetes more centralized agent-identity platforms will likely come down to how well it holds up against real enterprise agent sprawl — something not yet demonstrated publicly.
What's Next
Outerlimit says it will use the funding to scale product development and engineering, invest in customer success, and expand go-to-market operations across the United Kingdom and North America. No general-availability date or detailed product roadmap has been disclosed.