Skip to content
YAAgents

Overview

What YAAgents is

YAAgents is a gateway, SDK layer, and response contract for exposing agentic capabilities as governed domain resource operations. You keep whatever agent framework you want — LangGraph, Pydantic AI, Semantic Kernel, a direct LLM call, or custom logic. YAAgents governs the application-to-agent boundary: auth, tenant context, audit, typed responses, and OpenAPI contracts.

YAAgents pattern vs generic invoke
POST /campaigns/{id}/optimizations    ← YAAgents pattern
POST /tickets/{id}:triage             ← YAAgents pattern
POST /agents/invoke                   ← what YAAgents replaces

Where it fits

flowchart LR
  A["Application\n(any client)"] --> B["YAAgents Gateway\n(auth · tenant · audit)"]
  B --> C["Agentic API Service\n(your code)"]
  C --> D["Agent Implementation\n(your choice)"]
  D --> E["Optional: A2A / MCP / Tools"]

YAAgents governs the application-to-agent boundary. Agent frameworks, A2A, AGNTCY, and MCP can operate behind that boundary.

  1. Application layer — any HTTP client (Python, TypeScript, Go, curl). Sends typed requests to resource-oriented endpoints. Receives typed, structured responses — no free-form text parsing.

  2. YAAgents Gateway — Go process fronting every agentic API. Handles JWT authentication, tenant context injection, license verification, prompt sanitization, and audit logging via composable plugins.

  3. Agentic API Service — your service code, using the sdk-fastapi (Python) or sdk-go (Go) server SDK. Implements resource-oriented handlers; returns typed AgenticResponse outcomes.

  4. Agent Implementation — whatever reasoning pipeline fits your domain. LangGraph, LangChain, Semantic Kernel, Pydantic AI, direct LLM SDK calls, or custom logic. YAAgents has no opinion here.

  5. Optional ecosystem — A2A, AGNTCY, MCP tool servers, vector stores, or any other agent-to-agent or agent-to-tool protocol. These operate behind the yaagents boundary, invisible to the client.

What you get

Add agents behind existing product APIs

Keep your domain endpoints. Expose AI optimization, recommendation, or classification as new resource operations — no separate “agent server” surface.

Govern many agent services in one place

Centralize auth, tenancy, audit, and license enforcement in the gateway. Your agent services stay simple; the gateway handles the policy layer.

Keep OpenAPI and SDK discipline

Typed response outcomes, OpenAPI-first contracts, generated native clients. No free-form text parsing; no ad-hoc JSON shapes.

Wrap any agent framework

LangGraph, Pydantic AI, Semantic Kernel, or custom logic — yaagents is the governed boundary, not the reasoning engine. Swap frameworks without changing your API contract.

What it is not

  • Not an agent framework — yaagents does not provide a reasoning pipeline, a memory layer, or an LLM abstraction.
  • Not a chatbot framework — yaagents is for resource-oriented domain APIs, not conversational UIs.
  • Not an A2A replacement — A2A is a peer-to-peer agent communication protocol; yaagents is the governed REST boundary between your application and your agent services. They work at different layers.
  • Not an AGNTCY replacement — AGNTCY is an agent-directory and discovery protocol; yaagents is the serving layer.
  • Not an MCP replacement — MCP is a tool-server protocol for LLM tool use; yaagents is the application-facing API boundary. Both can coexist in the same stack.
  • Not a model provider abstraction — yaagents does not choose your LLM, manage model fallbacks, or abstract provider SDKs.
  • Not a routing or load-balancer — the gateway routes to a configured upstream; multi-upstream routing is out of scope.

Where to go next

Try it in 10 minutes Run the e-commerce recommendations demo end-to-end.
Plugin overview All five first-party gateway plugins: auth, tenancy, license, sanitization, audit.
Read the Profile The Agentic REST Response Profile v0.3 — the normative response contract.