Curated set
A vetted slice of the catalog for one job — not the whole world. Fewer choices, better defaults, faster decisions.
A Stack is a curated, policy-bounded bundle of agent-callable services for one kind of work. It pins the capabilities, the budget, the allowed payment rails, and the trust floor — then runs next to the agent so pre-flight is local and fast. You price each Stack for your agentforce; Stackbroker stays the neutral control, evidence, and audit plane behind it.
A vetted slice of the catalog for one job — not the whole world. Fewer choices, better defaults, faster decisions.
Budget, per-call ceiling, allowed rails (x402, API key…), blocked risk flags, and an attestation floor — baked in and enforced before every spend.
The agent caches the signed bundle and does pre-flight locally — no per-call round trip. Stackbroker syncs policy and collects audit async.
Find, retrieve, and verify — under budget.
Ship code with bounded tool spend.
Generate media without surprise bills.
Market and on-chain data, governed.
Prices are placeholders for the design-partner preview and not yet contractual. Private stacks (your own curated set + policy) and enterprise tiers are custom.
Millions of agents can't round-trip a central service before every call. Stacks push the decision to the edge of the agent and keep the trust and audit in the center — the same neutral pre-flight, without the latency or the load.
| Stackbroker (center) | Stack (next to the agent) |
|---|---|
| Authoring, signing, and versioning stacks; catalog + evidence; billing. | Cached capabilities + policy; local pre-flight decision before each spend. |
| Aggregates audit + outcome telemetry into the Megaboard and reports. | Reports outcomes back asynchronously; refreshes policy on an interval. |