---
title: LLM Applications, RAG & Agents
description: Orchestration frameworks, agent SDKs, structured output, gateways, and guardrails for building LLM applications.
tags: [machine-learning, llm]
---

# LLM Applications, RAG & Agents

Fetch when the target orchestrates LLM calls: RAG pipelines, agents, structured output, gateways, or guardrails. Each entry is one line and a documentation entry point; fetch install commands and API details from the entry point, never from memory. No entry is a recommendation. Embedding retrieval backends (ANN indexes and vector databases) live on the [vector-search](vector-search.md) page.

## Tools

| Tool | One line | Docs |
|---|---|---|
| LangChain | composable LLM applications and integrations | <https://docs.langchain.com/> — llms.txt: <https://docs.langchain.com/llms.txt> |
| LangGraph | stateful agent graphs in the LangChain family | <https://langchain-ai.github.io/langgraph/> — llms.txt: <https://langchain-ai.github.io/langgraph/llms.txt> |
| LangSmith | tracing, evaluation, and observability for LLM apps | <https://docs.langchain.com/langsmith> — llms.txt: <https://docs.langchain.com/llms.txt> |
| LlamaIndex | data framework for RAG and agents | <https://developers.llamaindex.ai/> — llms.txt: <https://developers.llamaindex.ai/llms.txt> |
| Haystack | pipeline-based LLM and retrieval framework | <https://docs.haystack.deepset.ai/> — llms.txt: <https://docs.haystack.deepset.ai/llms.txt> |
| DSPy | declarative LLM programs with automatic prompt optimization | <https://dspy.ai/> — llms.txt: <https://dspy.ai/llms.txt> |
| Semantic Kernel | Microsoft's model-agnostic agent SDK | <https://learn.microsoft.com/en-us/semantic-kernel/> |
| AutoGen | multi-agent conversation framework | <https://microsoft.github.io/autogen/> |
| CrewAI | role-based multi-agent teams | <https://docs.crewai.com/> — llms.txt: <https://docs.crewai.com/llms.txt> |
| PydanticAI | type-safe agent framework from the Pydantic team | <https://pydantic.dev/docs/ai/> — llms.txt: <https://pydantic.dev/docs/ai/llms.txt> |
| LiteLLM | one client and proxy over many LLM providers | <https://docs.litellm.ai/> — llms.txt: <https://docs.litellm.ai/llms.txt> |
| Instructor | structured output from LLMs via Pydantic | <https://python.useinstructor.com/> — llms.txt: <https://python.useinstructor.com/llms.txt> |
| Guardrails AI | input/output validation for LLM apps | <https://guardrailsai.com/guardrails/docs> |
| NVIDIA NeMo Guardrails | programmable rails for LLM conversations | <https://docs.nvidia.com/nemo/guardrails/> — llms.txt: <https://docs.nvidia.com/llms.txt> |
| Model Context Protocol | open protocol and SDKs connecting LLMs to tools and data | <https://modelcontextprotocol.io/> — llms.txt: <https://modelcontextprotocol.io/llms.txt> |

## Gotchas

- LLM-app docs sites publish llms.txt unusually often — probing for a rendition is especially worth it in this domain.