Open Model Access
OMA is a self-hosted-friendly AI inference gateway: drop-in OpenAI API compatibility, multi-provider routing, and native crypto payments for developers and autonomous agents.
Open Model Access
One OpenAI-compatible API across 21 curated models and multiple providers, with circuit-breaker protection and configured fallbacks where available.
Crypto-native billing
Deposit USDC on Base, or settle via x402 to credit a balance, then spend with an API key. Card checkout is optional when Polar is enabled.
Minimal OMA retention
SIWE wallet auth and magic-link email. OMA meters billing metadata but does not persist prompt or completion content; inference providers process requests as described in Privacy.
How OMA works
OMA is a single OpenAI-compatible endpoint —https://www.oma-ai.com/api/v1 — in front of a rotating catalog of chat models routed across multiple upstream providers. Each requested model resolves to its configured upstream route. Circuit breakers stop repeated calls to unhealthy routes, and mapped models can use a configured fallback where one is available.
You pay from a prepaid credit balance. Top up with a card when Polar checkout is enabled, deposit USDC directly on Base to the deposit address, or fund the balance via x402 (EIP-3009). An agent signs a payment authorization, OMA verifies and broadcasts the transfer, then settlement credits the account for API-key spending. There are no subscriptions; funding methods have their own deposit limits.
Auth is passwordless by default: magic-link email or SIWE with your existing wallet. Usage and spend are metered for billing; prompt and completion content are not logged.
What OMA is for
Migrate an OpenAI-compatible app
Keep your existing SDK and request shape. Change the base URL, use an OMA key, and choose a catalog model per request.
Migration quickstartCompare models without rebuilding
Use one catalog to compare model price, context window, provider, and capabilities instead of maintaining separate vendor integrations.
Browse modelsRun automated agent workloads
Use API keys with prepaid credits, direct USDC deposits, or the optional x402 settlement flow for software-driven funding.
Agent payment docsOperational boundaries
- Availability changes. Model capacity and upstream health are live conditions. Check the catalog and status page before depending on a model in production.
- Free chat is not free API traffic. The OMA-AI browser-chat allowance has daily caps; API-key calls spend credits at the listed rate.
- Inference still has a data path. OMA does not persist prompt or completion content, but the inference provider that handles a request must process that content to return an answer. See Privacy.