AI agents are getting smarter and soon they’ll be booking flights, submitting forms, and signing contracts on our behalf. But how can a payment provider, merchant, or financial institution verify that an AI agent is truly acting on behalf of a legitimate user?
And how can they know what the agent is actually authorized to do?
Vouched is introducing a new identity layer for AI agents: MCPI, the identity extensions to Anthropic’s MCP protocol. In this session, CEO Peter Horadan will walk us through how "Know Your Agent" works and what it means for the future of identity, delegation, and trust.
(00:00) – Intro
(03:35) – Why AI agents are exciting but risky
(06:25) – The security problem with giving agents your credentials
(10:33) – What’s missing from MCP: Identity and delegation
(12:00) – Five problems we must solve for agent identity
(16:30) – How MCP-I works: Agent shows up, user grants permissions
(19:31) – Demo
(30:12) – Why developers haven’t focused on identity for agents
(31:04) – Real-world privacy risks with current agent behavior
(32:36) – Can existing IAM systems adapt to agent delegation?
(33:45) – Role of reusable credentials and digital IDs
(40:03) – Do higher-assurance use cases need added safeguards?
(41:41) – Stopping agents from rotating DIDs to avoid detection
(43:59) – Peter’s advice to enterprise CTOs
(46:24) – Audience Q&A
(56:00) – AI agents will force every business to rethink workflows
📚 EXPLORE:
Dock Labs - https://www.dock.io/
Vouched - https://vouched.id
👨👩👧👧 FOLLOW:
LinkedIn - https://www.linkedin.com/company/docknetwork/
