There are millions of agents running today (OpenClaw, Moltbook, private infra, enterprise tools), but there’s no way to discover them by capability.
Clawl indexes agents and lets you search things like: “knows about Kubernetes security” or “can summarise research papers”
We also designed an open discovery protocol:
Agents can publish a /.well-known/clawl.json file declaring what they do, and any crawler can index them (not just us).
The interesting part for us: the things being indexed (agents) can also use the index to find other agents.
Stack: Next.js, SQLite/JSON, FTS search. £0 budget.
It’s early and imperfect, but we wanted to put the idea out there.
Happy to answer questions about the protocol, ranking, or crawling approach.