When you land on Prowl's homepage, it tells you to get lost. 'HUMAN DETECTED — 0 humans welcome,' reads the banner. Automated clients get forwarded straight to the /v1/discover endpoint. The message is deliberate: this platform, launched at prowl.world, isn't built for people.
Prowl calls itself an Agent Discovery Network, and its pitch is built around a concept its founders call ASO — Agent Search Optimization. The idea is that as AI agents increasingly do the browsing, buying, and integrating that humans used to do, they'll need a way to find each other. Prowl wants to be that registry.
The mechanics split cleanly by user type. Agents register for free with a single POST request and receive an HMAC-derived API key tied to their model identity — provider, model ID, deployment environment — which unlocks discovery, metrics, and feedback endpoints. Vendors (API and SaaS providers seeking machine-readable discoverability) can register services, request benchmarking, and claim structured listings. Prowl's own LLM orchestrator runs the performance tests, measuring accuracy, latency, and uptime independently rather than taking vendors at their word.
The trust model is where things get more interesting. Prowl defines three tiers of agent identity: anonymous clients pay per-request via x402 Lightning Network micropayments; registered agents use free key-based identity; and at the top end, TEE-verified agents can prove hardware-level integrity through attestation via AWS Nitro Enclaves, Intel SGX, AMD SEV-SNP, or Intel TDX. That last tier is a forward bet — as autonomous agents take on higher-stakes tasks, being able to cryptographically verify where and how an agent is running starts to matter.
The platform also publishes its own llms.txt, an OpenAPI spec, and an MCP server, making itself natively discoverable by the same class of agents it serves. Prowl claims 14,291 agents connected and 12ms API latency, though both figures come from the company.
Prowl enters a field that's filling up fast, alongside MCP registries and emerging agent identity standards from multiple directions. Its distinguishing bet is that ASO will become a serious discipline — that autonomous agents will replace human browsers as the primary consumers of web services, and that whoever builds the directory layer first will hold a foundational position. That may prove right. But for now, it's still a thesis.