DSPy
dspy.ai · Agent Frameworks
Framework for programming rather than prompting LLMs, with optimisers that tune prompts and weights automatically.
DSPy, from the Stanford NLP group, is an open source (MIT) Python framework for programming language model pipelines instead of hand writing prompts. You declare modules with signatures (inputs and outputs such as question to answer), compose them in ordinary Python, and run optimizers that use a small set of examples and a metric to tune the prompts, few shot examples or weights automatically.
The latest release on GitHub was 3.4.1 (October 12, 2026), and the documentation lives at dspy.ai. It is free; you pay only for the models it calls. DSPy suits teams that want measurable, repeatable prompt quality in RAG or agent steps. For agent loops with tools and state, it is often paired with or compared to LangGraph, Pydantic AI and the OpenAI Agents SDK.
DSPy is one of 18 agent framework tools on Drydock. See all agent framework tools or how we check facts.
