DSPy

An open-source Python framework from Stanford NLP for programming rather than prompting language models: you declare typed Signatures and use Modules and Optimizers to compile programs against a metric. Actively developed (v3.4.1, October 2026).

Website

https://dspy.ai

Who it's for

Teams who want to optimise prompts and pipelines systematically and testably, rather than hand-tuning them.

Pricing

Free and open source (MIT). Model API costs billed separately.

Category

Agent frameworks & orchestration

Last verified

2026-10-22 (current)

Confidence

high

Key facts

• Python 3.10+
• MIT licence
• Signatures, Modules and Optimizers (e.g. GEPA) for automatic prompt compilation
• Supports tool use and multi-step agents (e.g. ReAct)
• v3.4.1 released 12 October 2026

Sources

Every claim above is drawn from these; we do not publish unsourced specifications.

• https://dspy.ai/current/
• https://github.com/stanfordnlp/dspy

Notes & uncertainties

No commercial tier or hosted platform.

Something changed or out of date? Tell us — corrections are how this stays worth reading.