New AI Agent Identity framework from the AGNTCY

Securing the future of autonomous agents

Artificial intelligence (AI) agents are no longer just experimental technologies confined to research labs; they're now integrated into real-world workflows, transforming industries like finance, healthcare, and software development. Whether recommending a diagnosis in hospitals, automating HR processes, or analyzing customer feedback, autonomous agents operate independently, make decisions, and complete complex tasks.

However, the rise of these agents is also raising serious concerns: Are they secure? Can we trust them with our data and permissions? How do we ensure they perform as expected? The fact is, traditional identity solutions are unable to keep pace with agents: They act independently, can spin up or down in seconds, will cross organizational boundaries in order to achieve their outcome, and can easily adapt their behavior to different situations. 

Gone are the days of securing simple, linear workflows. Ensuring that each AI agent operates within its scope, reliably and securely, is critical. That's why AGNTCY Agent Identity was created.

Agent Identity from AGNTCY, an open source collective building the Internet of Agents, is a new software offering an identity framework that assigns, verifies, and manages credentials for AI agents. We'll explore the features it offers and how the framework secures AI agent workflows against potential risks.

New AGNTCY Agent Identity framework

AGNTCY Agent Identity establishes a secure and verifiable identity from the moment an agent is created and deployed. By assigning secure, trusted, and verifiable identities, AGNTCY ensures that every AI agent can be authenticated, tracked, and trusted before taking any action. The platform creates the necessary infrastructure for secure AI agent workflows that align with modern security standards.

Through collaborations with leading partners such as Cisco Duo, Skyfire, and Permit.io, AGNTCY is building trust in autonomous agents, and ensuring seamless onboarding, auditability, and governance across platforms, organizations, and ecosystems.

“AI agents are soon going to number in the billions. Cisco Duo is now working directly with Outshift to help AGNTCY push the frontiers of adoption in this new space, developing the frameworks and guardrails necessary for secure, productive adoption of agentic AI throughout the industry. We hear from our customers how important AI agent identity, policy, accountability, and auditing are to their AI roadmaps—and we're here to help accelerate that.”— Matt Caulfield, Vice President of Product, Identity at Cisco

Laying the foundation: Verifiable identity at agent registration

In this first release, we’re focused on the foundational step: assigning cryptographically verifiable identities to agents as part of agent onboarding and registration. This ensures that every agent can be authenticated, traced, and trusted before it takes action.

“Agent identity is the missing first step and a fundamental requirement that unlocks many downstream workflows any autonomous system must complete. Skyfire is partnering with Outshift to codify the industry playbook for those identity flows, so billions of verified agents can seamlessly access what they need.” — Ankit Agarwal, Head of Engineering at Skyfire 

AGNTCY supports both conventions and standards for agent identity assignment. Rather than prescribing a single method or format, we support an open model. This flexibility enables the integration of heterogeneous identity types while ensuring common ways to present and verify agent identifiers.

This includes identities issued by common identity providers (e.g., Okta, Microsoft AD, Auth0), as well as conventions like Google’s Agent2Agent (A2A) protocol with its standardized Agent Card. On the standards front, AGNTCY supports W3C Decentralized Identifiers (DIDs)—enabling open, decentralized, and verifiable agent identity using verifiable credentials (VC) for provenance, capabilities, and trust signaling.

Additional features include:

• Tamper-proof identity supported by signed ID badges that use asymmetric keys.
• Provenance you can trust using credentials that are anchored in a secure, tamper-proof trusted identity node (ledger).
• Dynamic and scalable authentication providing fast identity resolution across APIs, clouds and organizations.
• An open CLI that can generate keys, issues badges and register agents.

AGNTCY Agent Identity in action

AGNTCY's tools and trust components include the following key components:

• Identity CLI
  A command-line tool for users to create, manage, and verify agent identities. Handles key generation, badge issuance, registration, and credential resolution.
• Identity Node (IN)
  A secure, tamper-resistant identity ledger. Stores agent metadata, public keys, and credentials. Serves as the trust anchor for organizations and agents.
• ResolverMetadata
  A lightweight bundle containing an agent’s public key, Identity Node endpoint, and other metadata needed to automatically verify identity in real time.

Below is an example of how these components work together to establish agent identity and ensure each agent enters workflows securely, with minimal risk of unauthorized actions.

Step 1: Organization onboarding

An organization begins by setting up its identity infrastructure.

• A user opens the Identity CLI and generates an asymmetric key pair. The private key is securely stored (e.g., in HashiCorp Vault).
• The user registers their organization in the Identity Node.
• The Identity Node creates an organization entry and bonds the public key to it, establishing a root of trust.

Step 2: Badge creation and registration

A verifiable badge is created for a specific agent.

• The user provides an Agent Definition (e.g., OASF, A2A Agent Card, MCP Server) via the CLI.
• The CLI signs the definition with the private key, generating a ProofValue, ResolverMetadata, and a Verifiable Badge.
• These credentials are registered in the Identity Node, bound to the organization’s identity and public key.

Step 3: Badge verification

Anyone can verify the agent’s identity on demand.

• A user inputs the agent definition or agent ID via the CLI.
• The CLI fetches ResolverMetadata and the Verifiable Badge from the Identity Node.
• Using the agent’s public key, the CLI cryptographically verifies the badge, confirming that it was created and signed by a trusted party.

This process ensures every agent’s claims and identity are provable, allowing any system or user to verify authenticity before granting access or trust. This example flow showcases the first critical milestone: verifiable identity at creation. It's the foundation for scalable, secure, and auditable agent behavior.

Get started with AGNTCY Agent Identity

AGNTCY Agent Identity isn’t just a technical solution; it’s a forward-looking framework supporting innovation, trust, and governance as organizations scale their AI strategies. Organizations that adopt this framework will be able to block rogue or unauthorized agents before they act while establishing clear accountability by tracking every action back to specific agents. 

“We are excited to work with AGNTCY to help deliver the technologies and standards that will transform how developers manage permissions and access control in the Internet of Agents.” -Gabriel Manor-Liechtman, Vice President of DevRel, Permit.io

Read more about AI Agent Identity and how we plan to handle the dynamics of autonomous, ephemeral, and polymorphic agents. Or get involved today by joining our Agent Identity Working Group—an open forum for developers, security architects, researchers, and builders who believe trust must evolve alongside AI. Whether you’re hands-on with infrastructure or exploring the frontier of agent-based systems, your perspective matters.