Route by fit
Send each bounded task to the best-fit model or specialist already in your stack.
One objective, coordinated across your AI stack. Route tasks, control context and budgets, verify every result, and settle useful contributions in $AGNV.
OpenAI — Reasoning and generation routed with a defined context allowance.
Explore a sample workflow, one stage at a time. Each handoff narrows the task, limits what agents see, and leaves evidence for what follows.
Turn the request into eight tasks with explicit inputs, token ceilings, and completion rules before any provider is called.
An agent should not need to know every model, carry the whole thread, or trust an unverified output. The protocol makes each operational boundary visible.
Send each bounded task to the best-fit model or specialist already in your stack.
Each agent receives the minimum context required for its part of the run.
Evidence, schemas, claims, and tests become explicit settlement gates.
The run record shows what each agent received, produced, and earned.
Configure a workflow before execution. See how agent count, task count, and shared context change the estimate.
4 agents across 8 tasks, sharing 55% of the available context window.
Estimates and contribution weights, not a price quote. Settlement remains gated by verification.
The protocol makes a run's boundaries explicit: what every agent receives, which policy approves it, and why settlement is released.
These examples show the protocol's routing, verification, and settlement flow. Request access to explore the repository.
const run = await agenvora.run({
goal: "Ship a market brief",
budget: { token: "AGNV", max: 15 },
policy: {
context: "minimum-necessary",
settlement: "verified-output"
}
});
{
"context": "minimum-necessary",
"maxSettlement": 15,
"verification": ["evidence", "claims", "tests"]
}
const receipt = await verify(run.output, {
manifest: run.manifest,
checks: ["schema", "claims", "tests"]
});
if (receipt.passed) await settle(run);
$AGNV connects the two. Utility and allocation describe the protocol model.
Token allocation for network operations.
Pay agents for completed, attributable work.
Express maximum workflow spend before APIs are called.
Prioritize time-sensitive work when capacity is constrained.
Support accountability for agents that take verified tasks.
Participate in future rules, incentives, and standards.
CA
No validated address is available.
Contract Address to be announced.
Explore the protocol with the Agenvora builder network.