How intersystems-community/ai-hub-eap works
The AI Hub supports two principal workflows: developers build IRIS-native or LangChain applications that access governed AI resources, and IRIS owners expose business logic as MCP tools for external clients. The Config Store supplies configuration and credentials, while IRIS RBAC governs access; the LangChain extension also exposes IRIS Vector Search. The tree contains ObjectScript samples and small Python MCP examples, not the core SDK or server implementations, so several documented subsystems use enclosing directory paths and their links reflect README-level evidence rather than inspected source calls.
Main entry points: langchain LangChain integration · AI ObjectScript SDK · MCP MCP Server · ToolSet Toolset definitions
Components 18
AI SDK
- ObjectScript SDK
objectscript/cls/Sample/AI - LangChain integration
langchain - Vector Search
langchain
MCP publishing
- MCP Server
objectscript/cls/Sample/MCP - Toolset definitions
objectscript/cls/Sample/AI/ToolSet - IRIS business logic
objectscript/cls/Sample/AI
Configuration governance
- Config Store
langchain - IRIS Wallet
- IRIS RBAC
- Access auditing
Sample applications
- Agent examples
objectscript/cls/Sample/AI/Agent - Agent tools
objectscript/cls/Sample/AI/Tools - Addition MCP demo
langchain/addition_server.py - Multiplication MCP demo
langchain/multiplication_server.py
Other components
- Application developer
- External MCP client
- AI providers
- Remote MCP servers
Connections 19
- Application developer → ObjectScript SDK: builds with
- Application developer → LangChain integration: deploys with
- ObjectScript SDK → AI providers: abstracts access
- LangChain integration → AI providers: accesses models
- LangChain integration → Remote MCP servers: accesses servers
- LangChain integration → Config Store: reads configuration
- ObjectScript SDK → Config Store: finds configurations
- Config Store → IRIS Wallet: stores secrets
- ObjectScript SDK → IRIS RBAC: governs access
- LangChain integration → IRIS RBAC: governs access
- LangChain integration → Access auditing: supports auditing
- LangChain integration → Vector Search: exposes search
- External MCP client → MCP Server: invokes tools
- MCP Server → Toolset definitions: serves definitions
- Toolset definitions → IRIS business logic: exposes methods
- MCP Server → IRIS RBAC: governs access
- Agent examples → Agent tools: uses tools
- Addition MCP demo → External MCP client: serves addition
- Multiplication MCP demo → External MCP client: serves multiplication