How aws-samples/sample-agentic-platform works
Agentic Platform is a reference platform for running framework-diverse agents with shared core abstractions and gateway services. A user invokes an agent through its HTTP server; the server routes requests to a controller and agent, which returns a response or stream. The wider product includes chat, RAG, workflow, Jira and data agents; memory and retrieval gateways; and MCP servers. README documents AWS gateway-mediated access, but most agent internals and integrations were not sampled, so those links are shown only where documented or left out.
Main entry points: llm_gateway_client.py LLM Client · agentic_chat/server.py Chat HTTP · agentic_chat_controller.py Chat Controller · agentic_rag/server.py RAG HTTP
Components 33
Agent Applications
- Chat HTTP
src/agentic_platform/agent/agentic_chat/server.py - Chat Controller
src/agentic_platform/agent/agentic_chat/controller/agentic_chat_controller.py - Strands Chat
src/agentic_platform/agent/agentic_chat/agent/agentic_chat_agent.py - RAG HTTP
src/agentic_platform/agent/agentic_rag/server.py - RAG Controller
src/agentic_platform/agent/agentic_rag/controller/agentic_rag_controller.py - Strands RAG
src/agentic_platform/agent/agentic_rag/agent/agentic_rag_agent.py - Knowledge Tool
src/agentic_platform/agent/agentic_rag/tool/kb_tool.py - LangGraph Chat
src/agentic_platform/agent/langgraph_chat/chat_workflow.py - Workflow Controller
src/agentic_platform/agent/langgraph_chat/chat_controller.py - Jira Agent
src/agentic_platform/agent/jira_agent/jira_agent.py - Jira Controller
src/agentic_platform/agent/jira_agent/jira_controller.py - Glue Athena Agent
src/agentic_platform/agent/strands_glue_athena/agent_service.py - Data Controller
src/agentic_platform/agent/strands_glue_athena/agent_controller.py
Shared Core
- API Models
src/agentic_platform/core/models/api_models.py - Stream Models
src/agentic_platform/core/models/streaming_models.py - Memory Models
src/agentic_platform/core/models/memory_models.py - Request Middleware
src/agentic_platform/core/middleware/configure_middleware.py - LLM Client
src/agentic_platform/core/client/llm_gateway/llm_gateway_client.py - LiteLLM Client
src/agentic_platform/core/client/llm_gateway/litellm_gateway_client.py - Retrieval Client
src/agentic_platform/core/client/retrieval_gateway/retrieval_gateway_client.py
Gateway Services
- Memory Gateway
src/agentic_platform/service/memory_gateway/server.py - Memory Controllers
src/agentic_platform/service/memory_gateway/api/create_memory_controller.py - Memory Clients
src/agentic_platform/service/memory_gateway/client/memory/memory_client.py - Retrieval Gateway
src/agentic_platform/service/retrieval_gateway/server.py - Retrieve Controller
src/agentic_platform/service/retrieval_gateway/api/retrieve_controller.py - Vector Search Client
src/agentic_platform/service/retrieval_gateway/client/vectorsearch_client.py
MCP Servers
- Bedrock KB MCP
src/agentic_platform/mcp_server/bedrock_kb_mcp_server/server.py - Code Graph MCP
src/agentic_platform/mcp_server/code_graph_mcp_server/server.py - Graph Queries
src/agentic_platform/mcp_server/code_graph_mcp_server/tool/query_tools.py - Code Graph
src/agentic_platform/mcp_server/code_graph_mcp_server/tool/graph.py
Other components
- User
- Bedrock Knowledge Base
- LiteLLM Gateway Service
Connections 40
- User → Chat HTTP: invokes
README.md - Chat HTTP → Chat Controller: routes to
src/agentic_platform/agent/agentic_chat/server.py - Chat Controller → Strands Chat: invokes
src/agentic_platform/agent/agentic_chat/controller/agentic_chat_controller.py - User → RAG HTTP: invokes
README.md - RAG HTTP → RAG Controller: routes to
src/agentic_platform/agent/agentic_rag/server.py - RAG Controller → Strands RAG: invokes
src/agentic_platform/agent/agentic_rag/controller/agentic_rag_controller.py - Strands RAG → Knowledge Tool: uses
src/agentic_platform/agent/agentic_rag/agent/agentic_rag_agent.py - Knowledge Tool → Bedrock Knowledge Base: queries
- User → Workflow Controller: invokes
README.md - Workflow Controller → LangGraph Chat: runs
src/agentic_platform/agent/langgraph_chat/chat_controller.py - User → Jira Controller: invokes
README.md - Jira Controller → Jira Agent: invokes
src/agentic_platform/agent/jira_agent/jira_controller.py - User → Data Controller: invokes
README.md - Data Controller → Glue Athena Agent: delegates to
src/agentic_platform/agent/strands_glue_athena/agent_controller.py - Chat HTTP → Request Middleware: configures
src/agentic_platform/agent/agentic_chat/server.py - RAG HTTP → Request Middleware: configures
src/agentic_platform/agent/agentic_rag/server.py - Chat HTTP → API Models: uses
src/agentic_platform/agent/agentic_chat/server.py - RAG HTTP → API Models: uses
src/agentic_platform/agent/agentic_rag/server.py - Chat Controller → API Models: accepts
src/agentic_platform/agent/agentic_chat/controller/agentic_chat_controller.py - Chat Controller → Stream Models: emits
src/agentic_platform/agent/agentic_chat/controller/agentic_chat_controller.py - RAG Controller → API Models: accepts
src/agentic_platform/agent/agentic_rag/controller/agentic_rag_controller.py - RAG Controller → Stream Models: emits
src/agentic_platform/agent/agentic_rag/controller/agentic_rag_controller.py - Workflow Controller → Memory Models: uses
src/agentic_platform/agent/langgraph_chat/chat_controller.py - LLM Client → LiteLLM Client: delegates to
src/agentic_platform/core/client/llm_gateway/llm_gateway_client.py - LiteLLM Client → LiteLLM Gateway Service: requests
src/agentic_platform/core/client/llm_gateway/litellm_gateway_client.py - Retrieval Client → Retrieval Gateway: posts retrieval
src/agentic_platform/core/client/retrieval_gateway/retrieval_gateway_client.py - Memory Gateway → Memory Controllers: registers routes
src/agentic_platform/service/memory_gateway/server.py - Memory Controllers → Memory Clients: calls
src/agentic_platform/service/memory_gateway/api/create_memory_controller.py - Retrieval Gateway → Retrieve Controller: registers route
- Retrieve Controller → Vector Search Client: retrieves
src/agentic_platform/service/retrieval_gateway/api/retrieve_controller.py - User → Bedrock KB MCP: uses MCP
README.md - User → Code Graph MCP: uses MCP
README.md - Code Graph MCP → Graph Queries: calls
src/agentic_platform/mcp_server/code_graph_mcp_server/server.py - Graph Queries → Code Graph: queries
src/agentic_platform/mcp_server/code_graph_mcp_server/tool/query_tools.py - Strands Chat → LLM Client: uses gateway
README.md - Strands RAG → LLM Client: uses gateway
README.md - Jira Agent → LLM Client: uses gateway
README.md - Glue Athena Agent → LLM Client: uses gateway
README.md - Strands Chat → Memory Gateway: uses gateway
README.md - Strands RAG → Retrieval Gateway: uses gateway
README.md