Framework integrations
Agent-Gantry is designed to be the shared tool layer below many agent frameworks. These guides show how to expose the same trusted tool registry, retrieval strategy, and LLM routing policy from basic prototypes through production-grade agent systems.
LLM SDKs
- OpenAI / Azure / OpenRouter
- Anthropic Claude
- Google GenAI
- Google Vertex AI
- Groq
- Mistral via OpenAI-compatible endpoint
Frameworks
- Microsoft Agent Framework
- LangChain
- LangGraph
- LlamaIndex
- CrewAI
- AutoGen
- Semantic Kernel
- Google ADK
- Pydantic AI
- OpenAI Agents SDK
- Smolagents
- Haystack
- Agno
- Strands Agents
- DSPy
Protocols and storage
- MCP server/client routing
- A2A server/executor
- LanceDB persistence
- Qdrant / Chroma / pgvector adapters
- Nomic / OpenAI / sentence-transformers embeddings
- Cohere / cross-encoder rerankers
Framework-by-framework guides
Each guide is organized as a two-page learning path in one place: a basic page for setup, registration, and a first tool call, followed by an advanced page for dynamic tool selection, LLM calls, observability, and deployment patterns.
Microsoft Agent Framework
Basic usage, advanced orchestration, and LLM-call guidance.
LangChain
Basic usage, advanced orchestration, and LLM-call guidance.
LangGraph
Basic usage, advanced orchestration, and LLM-call guidance.
LlamaIndex
Basic usage, advanced orchestration, and LLM-call guidance.
CrewAI
Basic usage, advanced orchestration, and LLM-call guidance.
AutoGen
Basic usage, advanced orchestration, and LLM-call guidance.
Semantic Kernel
Basic usage, advanced orchestration, and LLM-call guidance.
Google ADK
Basic usage, advanced orchestration, and LLM-call guidance.
Pydantic AI
Basic usage, advanced orchestration, and LLM-call guidance.
OpenAI Agents SDK
Basic usage, advanced orchestration, and LLM-call guidance.
Smolagents
Basic usage, advanced orchestration, and LLM-call guidance.
Haystack
Basic usage, advanced orchestration, and LLM-call guidance.
Agno
Basic usage, advanced orchestration, and LLM-call guidance.
Strands Agents
Basic usage, advanced orchestration, and LLM-call guidance.
DSPy
Basic usage, advanced orchestration, and LLM-call guidance.
Recommended integration model
Start with the shared registry
Register Python callables once in Agent-Gantry, attach metadata, and export framework-specific tool objects only at the boundary of the selected framework.
Keep LLM calls behind adapters
Use the framework for conversation state and agent loops, then route model calls through your provider policy so usage tracking, cost controls, and fallback rules remain consistent.
Add advanced controls last
Introduce semantic tool retrieval, reranking, MCP/A2A executors, and tracing after the basic path has stable tests.
Microsoft Agent Framework
The first-class integration path. Use GantryContextProvider for per-run or per-call semantic tool injection, or GantryToolBridge when the tool set is fixed at construction. See the basic guide and advanced guide for orchestration, middleware, and AF 1.6 concurrent-workflow notes.