For customers
Customers select Auto, Frontier, or Fusion as their routing profile. Community capacity is a separate account permission for an approved execution mode.
Community inference marketplace
Match approved AI work with reviewed provider nodes, under explicit customer consent.
Customers define where work may run. Providers declare models and capacity; the router admits concurrency after measurement, while authenticated polling establishes current availability.
Customers select Auto, Frontier, or Fusion as their routing profile. Community capacity is a separate account permission for an approved execution mode.
Offer selected local models through an outbound worker. Keep concurrency, model residency, and availability under local control.
Route only to eligible nodes, verify scoped result signatures and, for Fusion, require an independent verifier's successful response. Settlement records stay separate from prompt content.
Agentic commerce direction
TEKIZ.AI is targeting x402-compatible quotes, payment authorization, and Bazaar discovery so an agent can find eligible inference, compare a bounded price, pay programmatically, and receive an OpenAI-compatible result.
x402 payments are planned. Paid Community Direct offers are currently disabled. They can launch after a reviewed payment rail binds authorization to the customer, API key, client peer, provider peer, tariff, and expiry.
Publish machine-readable inference resources through x402 Bazaar metadata and TEKIZ.AI's model catalog.
Return a route-specific ceiling before execution, with model, execution class, residency, and expiry attached.
Target x402's HTTP-native payment handshake, including usage-based authorization where the reviewed scheme supports it.
Optimize price only after consent, capability, security class, residency, availability, and quality requirements are satisfied.
Capacity evidence confirms available compute; it does not protect request contents from the node operator. Each execution class states who can see plaintext so customers can consent to that exact boundary.
Not yet available
A consenting Free or Lite client is designed to connect directly to a Community peer over QVAC's encrypted transport. TEKIZ.AI does not see the content, but the provider processes plaintext. Offers remain disabled until a reviewed payment authorization rail is implemented.
Private alpha for an explicitly electing Pro tenant. Reviewed server capacity presents evidence, but its operator may access request plaintext.
Planned confidential execution requires fresh remote attestation bound to the client session. This is the only Community class that can claim confidentiality from the infrastructure operator.
Availability alone never creates an assignment. Every request passes a sequence of customer and provider boundaries.
The durable account policy allows the exact Community execution mode and residency boundary.
An active, recently polling, operator-attested node has an admitted model and available slot.
For router-mediated Fusion, an independent verifier owns the final customer response.
Only work used in a successful verified response becomes settlement-eligible.
The economic layer follows verified work, not idle time or advertised hardware.
Rates and automatic payouts are not live. Private-alpha settlement remains manual while agreements, tax handling, fraud controls, refunds, and disputes are completed.
Failed, late, duplicate, cancelled, rejected, or unused results are not settlement-eligible.
The draft closed beta targets 70% of net Community inference revenue for providers whose work is accepted.
Planned credits offset Pro or Max routing first, then eligible Community inference.
Task fit, live capacity, safe residency, router-observed speed, and accepted outcomes guide future assignments.
Operators can prepare a reviewed node. Customers can set an explicit Community consent boundary in their account.