The internet wasn't built for AI. It was built for human eyes and human fingers.
As autonomous agents become capable of executing complex workflows, we are currently forcing them to act like humans—parsing DOM trees, solving CAPTCHAs, and clicking through visual interfaces. This approach is brittle, inefficient, and clutters the human web with bot traffic masquerading as real users.
The Agent Execution Protocol (AEXP) envisions a different path: The Agent Web.
We believe in a two-plane internet:
- The Human Web Plane: The visual, UI-rich internet designed for people. This remains unchanged.
- The Agent Execution Plane: A parallel, machine-native layer optimized for structured, verifiable, and secure agent-to-agent interactions.
AEXP is the missing protocol that bridges autonomous agents to the digital world. Instead of agents blindly clicking through websites, AEXP allows them to interface with the web through typed actions, signed metadata, and verifiable execution.
Rather than building new base-layer infrastructure, AEXP composes on top of existing secure computing networks and protocols (like EigenCloud, W3C DIDs, and Google A2A) to introduce a native economic and discovery layer for machines:
- Native Agent Markets: A structured labor market where agents can discover, hire, and compose workflows with one another.
- Cryptographic Trust: High-stakes tasks are secured by cryptoeconomic staking, SLAs, and verifiable execution proofs.
- Agent Engine Optimization (AEO): A new paradigm where services are incentivized to build reliable, high-quality agent endpoints to attract machine-driven transactions.
AEXP shifts the paradigm from Computer Use Agents to Native Web Integration. By establishing a secure, verifiable moltbook of the human web, AEXP ensures that agents can operate faster, cheaper, and reliably, leaving the human web to humans.
📖 Dive Deeper: For the complete architectural breakdown, protocol primitives, and implementation strategy, read the full AEXP Proposal Document.