Regis AI Query API
Natural-language question endpoint (POST /v2/ai/query) that runs a server-side agent over the same tool set as the MCP server and streams the result as Server-Sent Events — status, tool_call, tool_result, text, suggested_actions and done frames — in `ai` mode (written answer plus data) or `data` mode (raw tool JSON only). Stateless, one question per call, query up to 4,000 characters; quota counters returned in X-Quota-* headers with 429 + Retry-After on exhaustion. Marketed as Regis.