Ownware
Own Your AI (OYA) · API

Own Your AI (OYA) API, as shipped in the download

HTTP API

oya serve exposes 18 routes. Everything except /health and the static app is token-gated: the engine fails closed, so an unauthenticated request is refused rather than served.

RouteTokenWhat it is for
/adminyesAdmin surface for the web app: config, connectors, skills.
/admin/configyes—
/admin/mcpyes—
/admin/modelyesModel wizard — hardware detection and recommendations.
/admin/model/downloadyesStarts and reports a model download.
/bridgenoBridge status.
/bridge/startnoStarts a configured bridge.
/deliverableyesWhat the current session actually produced, as distinct from what it said.
/deliverablesyesThe full list of deliverables.
/healthnoLiveness. The only route deliberately open, so a supervisor can poll without a token.
/ledgeryesThe egress ledger — every host this process contacted.
/mcpyesThe MCP endpoint: this machine's tools, exposed for another agent to call.
/model/noProxies model requests from the browser, so the page never holds your endpoint credentials.
/sessionyesSession history and management.
/stateyesCurrent engine state for the web app.
/v1noOpenAI-compatible surface.
/v1/chat/completionsnoOpenAI-compatible chat completions, proxied to your local model.
/v1/modelsnoModels available from the configured endpoint.

The token is written to ~/.oya/serve-token and stamped into the page the engine serves, so the browser app authenticates without you copying anything. See [Running the engine](../guides/engine.md).


_Generated from cli/oya.mjs._

← Back to Own Your AI (OYA) · Manual · Quickstart · Test run

Affiliate program
Recommend tools people own — earn 35% on every sale. 90-day tracking, instant delivery, payouts by Lemon Squeezy.
Become an affiliate →