Shane Deconinck Trusted AI Agents · Decentralized Trust

With the rapid evolution of AI agents, standards are emerging for how they authenticate, connect to tools, and communicate with each other.

Interactive walkthroughs of the most important ones: what they solve, how they work, and where the gaps remain.

UserOBOAgentAgentGatewayCedar policiesMCPToolsA2AAgentsacross organizational boundariesTSPTrust Spanning Protocol

Everything above the line works within your trust domain. TSP is the emerging layer for when agents need to verify strangers.

OBO
How agents act on behalf of users Step-through demo of OAuth On-Behalf-Of: how agents act as users via RFC 8693 token exchange, delegation chains, and where OBO breaks at multi-hop.
→ Gateway
The control point for agent traffic How AgentGateway secures agent traffic: MCP federation, centralized OAuth, Cedar policy authorization, and governing agent-to-tool connections.
→ MCP
How AI agents connect to tools and data Step-through demo of Anthropic's MCP: architecture, OAuth authorization, security risks of tool integration, and what the protocol doesn't solve.
→ A2A
How agents discover and talk to each other Step-through demo of Google's A2A protocol: how agents discover each other, exchange tasks, and where cross-domain trust breaks down.
→ TSP
Trust across organizational boundaries How TSP enables trust between organizations without a shared authority: verifiable identifiers, credentials, and cross-org agent verification.
→