Routes OpenAI-style chat completions requests to Agent Y2 and streams text or client-side tool-call responses to compatible clients. v1 supports `messages`, `stream: true`, optional `model` set to `y2-agent` or `agent-y2`, optional `threadId`, optional bounded `metadata`, and bounded OpenAI function tools. Arbitrary models are unsupported. Without client tools, Agent Y2 keeps its fixed Copilot instructions and Y2 tools. A non-empty `tools` array selects harness mode: system and developer messages become the client instructions, tool-call history is preserved, and tool calls are returned to the client without server-side execution. v1 requires an API key and does not support x402. Both Agent Y2 routes share a throttle of 5 requests/minute and 100/day per key, plus 10 requests/minute and 250/day per user or workspace. A `402` indicates subscription, Agent Y2 credit, or upstream-provider credit exhaustion, not an x402 challenge.

Loading API reference...

POST /agent-y2/chat/stream