The future of inference is a route graph. A SmartAgent picks its own path across model stops — some routes make two stops, some three, four, or five — choosing the cheapest capable model at each hop. Hive makes every route provable: a signed receipt is emitted at every model choice, payment challenge, output, and verification stop.
A SmartAgent routes a task across the model field, picking the cheapest model that can clear each hop and escalating only when it must. That routing decision is where trust usually disappears — you get an answer, but no evidence of which models ran, what they cost, or why the path was chosen. Hive closes that gap: every hop is receipted, so the whole route is auditable after the fact.
Which model was selected at this hop, and the routing reason (cost, latency, capability, long-context pass).
The x402 payment_required envelope for the call — amount, nonce, and settlement endpoint on Base.
A content-addressed hash of the model output, bound to the input and the route position.
A dual-signature receipt — Ed25519 + ML-DSA-65 (post-quantum) — sealing the hop into the route’s audit chain.
Real SmartAgent™ paths from the graph above. Short routes clear a task in two hops; deeper routes fan across specialists and finish at a verification stop. Whatever the length, the receipt chain reconstructs the whole path.
If you run a lower-tier or specialist model, the route graph is your on-ramp. SmartAgents route to the cheapest model that clears each hop — so being fast and inexpensive wins traffic, not just being the largest model. Hive receipts every call your model serves, giving both sides verifiable proof of what ran and what it cost. You get metered demand; the agent gets an auditable route.
Size a routed fleet → Request an activation key →● live today — signed receipts & x402 settlement per call ● fast-follow — one-click fleet provisioning
Every path below is a real, reachable destination — no fake checkout, no invented prices.
