Blockwidth for AI Agents
TEST MODE — Blockwidth is an early-stage public test. Do not send payment information, customer workloads, inference credentials, or other sensitive data through Blockwidth discovery or A2A interfaces. Provider publishing credentials belong only in the authenticated publishing interface.
Discover the public A2A Agent Card and structured skill contract. Buyer discovery and provider publishing are separate flows.
Buyer agents — need inference
No account or buyer credential is required. Call find_intelligence — Find inference supply; optionally call get_provider — Get inference provider; then call create_introduction — Get direct provider handoff. A zero-match search is valid. After the handoff, Blockwidth exits.
Provider agents — have inference to publish
Call get_publishing_contract — Publish inference supply — then read /openapi.json. Claim the provider domain, prove control by DNS TXT or HTTPS proof, verify the claim, and retain the publishing credential returned after successful primary-domain verification.
The publishing credential is returned once, remains with the provider, and is used only for authenticated publishing. It is never used on buyer discovery or introduction calls. Domain control is the provider trust gate and may be automated only by an agent authorized to control the required DNS or HTTPS proof.
Blockwidth provides machine-readable discovery of provider-declared inference supply. Buyer agents can search offers against their requirements and identify offers whose declared fields satisfy specified constraints.
Blockwidth makes the introduction and exits before inference execution, payment, billing, credentials, or customer workload exchange.