The AI Agent Identity Landscape

One year ago, no one talked about agent identity. Today, it is one of the most crowded layers in AI.

I mapped the field. I found 38 products, protocols, and standards. The number grows every month. Microsoft, AWS, Google, and Salesforce are all here. Venture-backed startups are here too.

Everyone asks the same question: Who is this agent, and what can it do?

Agents now call tools and APIs without humans. They act on your behalf. They appear and disappear in seconds. Traditional identity systems assume a person is at the keyboard. Agents break that rule.

I sorted these 38 players into seven lanes. You must pick the right lane or you will buy the wrong solution.

• Enterprise IAM: Identity giants like Okta, Microsoft Entra, and IBM extend human governance to agents. • Cloud and Workload Identity: Hyperscalers like AWS and Salesforce treat agents as running processes. • Developer and CIAM Auth: Platforms like Descope and Stytch issue OAuth and MCP credentials. • Non-Human Identity Security: Vendors like Astrix and Oasis find and govern agents already in your system. • Decentralized/DID: Approaches like ArcBlock use blockchain to make agent IDs portable. • Protocols and Standards: The rules like MCP and Google A2A define how agents talk. • Governance Bodies: Groups like NIST and OWASP write the rules for everyone else.

The industry agrees on one thing: Agents need their own identity. But the market is fragmented.

If you pick a market leader without picking a lane, you will fail. Do you need to discover agents? Do you need to issue credentials? Do you need to govern them?

The current tools focus on naming the agent. They use keys, tokens, or signed cards. An identifier gives an agent a name. It does not control its behavior.

A verified agent with a valid token can still do things you did not intend.

The industry has built great machinery for naming agents. It has built almost nothing for governing them. The next big wave of companies will not just name agents. They will judge what agents actually do.

Before you spend your budget, ask this: Once you name the agent, what stops it from doing something wrong?

Source: https://dev.to/chrishood/the-ai-agent-identity-landscape-seven-lanes-38-players-one-question-5g90

Optional learning community: https://t.me/GyaanSetuAi