Keycard is Developing the Identity and Trust Layer for AI Agents

Keycard, building identity infrastructure for AI agents, led by Matt Creager, Jared Hanson, and Ian Livingstone, has secured a $38 million combined inception round led by Andreessen Horowitz, Acrew Capital, and Boldstart Ventures with participation from Mantis VC, Tapestry Ventures, Essence Ventures, Exceptional Capital, Modern Technical Fund, Vermillion Cliffs Ventures, and many incredible angels.

A new era of computing is emerging, where software evolves in real time as networks of AI agents act on behalf of people and businesses. Fueled by advanced large language models, cheaper computation, and smarter context-engineering, software can now adapt to new tasks on the fly, leveraging the prompts and tools available at any given moment.

Software is entering a fundamental shift: humans no longer need to write, review, or deploy code to unlock new capabilities. AI agents can now interact through voice, video, and text, creating contextual, embedded experiences that redefine how we work and engage with technology.

This change could spark a productivity wave even larger than SaaS, mobile, and cloud combined, moving repetitive cognitive tasks from humans to machines and freeing people to focus on higher-value work, much like the steam engine drove the industrial revolution.

In the fast-moving world of AI agents, traditional trust models built for humans are no longer enough. Agents are now making decisions, writing code, and taking actions on systems of record and commerce independently, without human intervention.

While large language models give agents the ability to handle complex cognitive tasks, their “superpower”, they also introduce risk. Hallucinations, logic gaps, and unpredictable actions can lead to serious consequences, from data loss to erroneous transactions, making the potential rewards weigh against the risks for many organizations.

Current AI agents are like 1990s cruise control: they can maintain the right pace, but a single mistake can have serious consequences, forcing humans to stay constantly alert and limiting the value of automation.

Transitioning to a truly agent-driven world requires new tools that make AI controllable, understandable, and aligned with human intent, backed by deterministic guarantees.

Every major computing wave has relied on advances in identity and access. Mainframes, the internet, and the cloud each introduced new ways to authenticate, authorize, and protect users. The AI agent era now demands the next evolution: traditional, human-focused IAM systems can’t support ephemeral, autonomous agents. Agents need dynamic provisioning, federated identity, attested runtime, and traceable delegations to operate securely at scale, yet today’s siloed machine identity systems prevent them from moving seamlessly across networks and applications.

AI agents turn long-standing machine identity challenges into a critical adoption barrier. Agents must act only within user intent, with access that can be instantly revoked, while platforms need full visibility into their actions.

Keycard provides the infrastructure to make this possible, replacing static roles and long-lived credentials with ephemeral, identity-bound tokens. These tokens support task-scoped policies, mixed delegation chains, and cryptographic tracking, giving developers and enterprises the tools to safely deploy and control AI agents at scale.

Built on years of experience in identity and access, Keycard is designed for the distributed, federated nature of AI agents, enforcing authentication and authorization at the edge. Unlike a gateway, it’s a secure token service that brokers credentials across applications, agents, and services, extending existing identity systems.

Agents receive tokens scoped to their exact task, preventing misuse of overscoped credentials or static secrets. Every issuance, delegation, and action is tracked in a contextual audit log, giving developers, users, and security teams full visibility into who used an agent, what tools were accessed, and which policies applied.

From experimentation to production, Keycard provides the infrastructure to build, adopt, and scale trusted AI agents with security, control, and confidence.