overlay: 1.0.0 info: title: API Evangelist enhancements for the Nexla GenAI (RAG + MCPaaS) API version: 1.0.0 extends: ../openapi/nexla-genai-mcpaas-openapi.json x-generated: '2026-08-26' x-method: generated x-source: >- Derived from the harvested spec at https://api-genai.nexla.io/openapi.json plus mcp/nexla-mcp.yml and mcp/nexla-tool-crosswalk.yml. The original spec is never mutated. actions: - target: $.info description: >- The harvested spec declares no servers[] block and no securitySchemes, so a generated client has neither a host nor an auth model. Record both from the documented deployment. update: x-api-evangelist-notes: base_url: https://api-genai.nexla.io mcp_endpoint: https://api-genai.nexla.io/mcp/service_key/{server_key} mcp_transport: streamable-http auth: 'Authorization: Bearer , or OAuth 2.0 for clients that cannot set headers' oauth_protected_resource: https://api-genai.nexla.io/.well-known/oauth-protected-resource docs: https://docs.nexla.com/dev-guides/mcp-tools - target: $.servers description: Add the production host, which the published document omits entirely. update: - url: https://api-genai.nexla.io description: Nexla GenAI / MCPaaS production host (probed 2026-08-26). - target: $.components.securitySchemes description: >- Declare the documented bearer service-key scheme. The published contract has an empty securitySchemes map even though every operation requires authentication. update: NexlaServiceKey: type: http scheme: bearer description: >- Permanent Nexla service key presented as `Authorization: Bearer `. Documented at https://docs.nexla.com/dev-guides/authentication/overview and https://docs.nexla.com/dev-guides/mcp-tools/authentication. - target: $.paths['/v1/toolsets/{tool_set_id}/exports/{export_id}:retire'].post description: Flag export retirement as terminal for agent consumers. update: x-agentic-consequence: destructive x-agentic-reversible: false x-agentic-note: 'Retiring an export takes an MCP server offline; pause is the reversible alternative.' - target: $.paths['/v1/tools/{tool_id}:execute'].post description: >- Mark the tool-execution operation as the consequence boundary — the effect depends entirely on the Nexset the tool was minted from, so it cannot be classified statically. update: x-agentic-consequence: variable x-agentic-note: >- Consequence is determined by the backing tool definition. Read GET /v1/tools/{tool_id}/definition before executing. Every call writes an audit receipt retrievable via GET /v1/receipts/{receipt_id}.