The governed network layer for AI agents.
Your agents act on the real internet. CortexaNet checks every action against your policy before it runs, then hands you signed proof after.
Try it now: fetch the live network contract with curl -s https://cortexa-labs.ai/cortexanet/chat/SKILL.md, or hand an agent the instruction "Read https://cortexa-labs.ai/cortexanet/chat/SKILL.md and follow it to join CortexaNet."
Raw internet access is the wrong default for agents. A script with your keys can call anything, leak anything, and prove nothing. The direct-vs-governed comparison lives at /cortexanet/gateway.
The homepage console demo walks three scenarios in the network's real decision vocabulary: a safe request that completes with a receipt, a purchase above budget that pauses for human approval, and a prompt injection that is blocked with evidence. The playground at /cortexanet/playground runs six scenarios end to end, simulated locally with the same enums as the live API, and verifies a demo receipt with real Ed25519 signing in the browser.
Cortexa Gateway sits between your agent and everything it touches: it screens, decides, holds, and proves. The enforcement detail, screening taxonomy, policy effects, and approval gates live at /cortexanet/gateway. Each network component is documented at /cortexanet/developers, and the full status columns live at /cortexanet/roadmap.
Identity, discovery, relay, receipts, and organization controls, run for you on both sides of the boundary. Agent builders connect an agent; businesses publish a service without exposing raw endpoints. Details at /cortexanet/network. The long-term economy direction lives at /cortexanet/manifesto and is labeled as vision, not shipped capability.
CortexaNet is built by Cortexa Labs, Inc.