The fundamental object isn't "an agent." The best answer to an intent may be a dedicated agent, a business API, an MCP server, or a marketplace — so AgentSort ranks Provider × Capability × Context. Sort outcomes, not labels.
One global score is not enough.
An agent that is exceptional at code generation may be poor at travel booking. AgentRank evaluates every provider in the context of a task and its constraints — pick a task and watch the signal weights shift.
The rank of provider p for task q under context c — with a recursive, PageRank-like authority component where every edge carries task, outcome, cost, latency, trust and time.
Organic and sponsored never mix.
From a capability universe to a dynamic toolset.
AgentSort Resolve turns a massive capability universe into a small, evidence-rich toolset a frontier model can reason over efficiently — assembled per intent, discarded after.
const result = await agentsort.resolve({
intent: "Find and book a reliable emergency
plumber in Brooklyn tonight",
top_k: 5,
constraints: {
verified: true,
available_within: "2 hours",
required_verification_level: 3
}
});
Discovery is not the hard part. Trust is.
The question is no longer “can I find an agent?” — it is “can I trust the thing I am about to let an AI invoke?” AgentVerify is a four-level ladder, independently assessed.
required_verification_level >= 3
Claims tell us what an agent says it can do.
Outcomes tell us what it actually does.
Registries tell AgentSort what providers claim. The Execution Graph records what actually happens when those capabilities are invoked — across platforms, not inside one.
Five linked graphs, one answer.
For this precise intent, under these constraints, which provider is most likely to successfully accomplish the task right now?
Organic ranking is never for sale.
AgentExchange is a separate commercial feed a host platform may choose to surface and label. Money can never improve organic AgentRank.