Topic hub · 18 claims
Agent frameworks — orchestration libraries for LLM apps
Frameworks that orchestrate LLMs in multi-step agent pipelines. Each picks different defaults for tool-use, memory, retrieval, and observability.
Why frameworks emerged
By mid-2022 the agent loop pattern — model emits tool call, runtime executes, model receives result, repeat — was clearly the production shape. Writing it from scratch for each project produced inconsistent error handling, inconsistent retries, inconsistent observability. Frameworks like LangChain (October 2022) and LlamaIndex (November 2022) emerged within weeks of each other to standardize.
The current landscape
As of 2026: LangChain (orchestration breadth) + LlamaIndex (retrieval-first RAG) dominate Python. DSPy (Stanford) offers programs-not-prompts. Pydantic AI brings type-safety. OpenAI Agents SDK + Anthropic SDK are vendor-native. Vercel AI SDK owns Next.js. Each has a different mental model — pick by archetype + audience + commitment level.
The cross-vendor convergence
Anthropic's Model Context Protocol (November 2024) is the cross-vendor standard for tool exposure. Adopted by Anthropic, OpenAI, and most major frameworks within ~6 months. The framework count may eventually drop as MCP absorbs per-vendor SDKs — but as of 2026 the seven-framework landscape is what production developers face.
Defined terms (3)
- Agent framework
- A library that orchestrates LLM tool-use loops, retrieval, memory, and observability. Examples: LangChain, LlamaIndex, DSPy.
- Tool-use loop
- The multi-turn pattern: model emits tool call, runtime executes tool, model receives result, model decides next step or final answer.
- Programs-not-prompts
- DSPy's paradigm: write structured programs (modules + signatures) that get optimized for prompts and few-shot examples rather than hand-writing prompts.
All claims in this topic (18)
- Anthropic Computer Use·publicly released on 2024-10-22 by Anthropic — Claude 3.5 Sonnet computer use beta(1.00 · 2 sources)
- Anthropic Tool Use (general availability)·publicly released on 2024-05-30 by Anthropic(1.00 · 2 sources)
- AutoGen·publicly released on 2023-09-25 by Microsoft Research(1.00 · 2 sources)
- DeepSpeed·publicly released on 2020-02-13 by Microsoft Research(1.00 · 2 sources)
- Haystack·publicly released on 2020-04 by deepset GmbH(1.00 · 2 sources)
- Instructor library·introduced in Jason Liu 2023 — structured outputs from LLMs via Pydantic(1.00 · 2 sources)
- JAX·publicly released on 2018-12-10 by Google Research(1.00 · 2 sources)
- LangChain framework·publicly released on 2022-10-25 by Harrison Chase(1.00 · 2 sources)
- LangSmith·publicly released on 2023-07-18 by LangChain — LLM observability + evaluation platform(1.00 · 2 sources)
- LlamaIndex framework·publicly released on 2022-11-09 by Jerry Liu (originally GPT Index)(1.00 · 2 sources)
- Microsoft Semantic Kernel·publicly released on 2023-03-17 by Microsoft(1.00 · 2 sources)
- Model Context Protocol (MCP)·publicly released on 2024-11-25 by Anthropic(1.00 · 2 sources)
- OpenAI Function Calling·publicly released on 2023-06-13 by OpenAI(1.00 · 2 sources)
- PyTorch·publicly released on 2017-01-18 by Facebook AI Research(1.00 · 2 sources)
- ReAct (Reasoning + Acting)·introduced in paper ReAct: Synergizing Reasoning and Acting in Language Models (Yao et al., 2022)(1.00 · 2 sources)
- TensorFlow·publicly released on 2015-11-09 by Google(1.00 · 2 sources)
- Toolformer·introduced in Schick et al. 2023 — self-supervised LLM tool-use(1.00 · 2 sources)
- CrewAI·publicly released on 2023-12 by João Moura — multi-agent orchestration framework(0.95 · 2 sources)