Research · October 31, 2026

Infisical's $40M Series B — Giving AI Agents Access Without Giving Them the Keys

Infisical raised a $40 million Series B at a $600 million valuation led by Elad Gil to secure AI agent access without handing agents the credentials.

What happened

On October 27, 2026, Infisical, a San Francisco-based security infrastructure company, announced a $40 million Series B funding round at a $600 million valuation. The round was led by Elad Gil, who previously invested in the company's Series A, with Gradient and Y Combinator among its other investors.

Infisical says it will use the new funding to accelerate product development, expand its engineering and go-to-market teams, and build out its security infrastructure for AI agents as enterprise adoption increases.

The problem Agent Vault is aimed at

Infisical's starting point is a shift in who — or what — logs in. Its platform already helps organizations manage secrets, certificates, encryption keys, and privileged access across their infrastructure. The new pressure, the company argues, comes from AI agents: to do useful work, an agent needs access to source code, databases, cloud infrastructure, and other critical systems, and that access runs on credentials — API keys, passwords, and authentication tokens.

Handing those credentials to the agent is, in the company's framing, the risk itself. A credential sitting inside an agent's environment can leak in a cyberattack or be sent, inadvertently, to a language-model provider or another tool the agent talks to. Infisical's answer, Agent Vault, is to give the agent the access without giving it the credential: companies define exactly which systems each agent may reach, and the agent performs its work inside that defined scope while the underlying secrets stay out of its hands.

Co-founder and CEO Vlad Matsiiako puts the principle in human terms: "You wouldn't give an employee unrestricted access to every system in your company, and AI agents shouldn't be any different." He adds: "Agents need access to company systems to do their jobs, but giving them actual passwords and API keys is a major security risk. We let agents get their job done safely without ever having those credentials in their hands."

Lead investor Elad Gil frames the bet in the same direction, calling AI agents "a new class of users inside every enterprise" and saying that securing their access "will be one of the biggest cybersecurity challenges of the next decade." He says Matsiiako and co-founders Maidul and Tony "recognized early that AI agents would create an entirely new set of security challenges" and that he believes the platform "can become essential infrastructure for the AI era" — an investor's expectation, not a measured outcome.

What the company already reports

All of the scale figures below are company-reported in the October 27 announcement, not independently verified. Infisical says it serves more than 1,300 customers and 300,000 developers, secures 10 billion secrets every day, and that its open-source project has earned more than 30,000 GitHub stars. Named customers in the announcement include Airbus, Volkswagen, Hugging Face, OpenRouter, Lucid, First American Financial Corporation, Wise, and Clay.

The pitch, read carefully, is narrower than "AI security" in general. Infisical is not claiming to judge whether an agent's actions are wise, only to control what the agent can touch and to keep the keys out of the agent's reach while it works. That is a deliberately boring-sounding control — least privilege, enforced mechanically — applied to a new kind of user that never sleeps, never forgets a credential it has seen, and can act across many systems in a single run. If agents become routine enterprise workers, the credential problem compounds with every deployment; if agent adoption stalls or stays narrow, the same product is still a secrets-management platform, but the Series B thesis weakens.

What this does not show

A $40 million round at a $600 million valuation is a price on an expectation, not proof of the category. No revenue figure is disclosed in the announcement, and no lifetime total raised is stated on a reconciled basis, so neither is claimed here. The 1,300-customer and 10-billion-secrets figures are company-reported totals with no breakdown by paying versus free users, no retention or churn data, and no audit of the secrets count alongside them. The named customers are selected by the company, and the announcement publishes no incident, breach-prevention, or credential-leak measurement that would let a reader test whether Agent Vault stops the failures it describes. The competitive field — established identity and secrets-management providers, and the cloud platforms' own access controls — is not addressed in the announcement at all. What would turn this round from a thesis into evidence is adoption measured the hard way: agents running in production under Vault controls, at named enterprises, with access scopes audited and credentials demonstrably never exposed — reported by customers, not only by the vendor.

Sources