Home

Welcome to the Nudge Wiki 👋

Don’t parse your agents. Nudge them.

Nudge is a typed, replayable, budget-aware programming language for LLM agents that compiles to Python and TypeScript.

This wiki is a home for practical guides, examples, and community knowledge to help you build with Nudge.

Start here

Visit the [project README](https://github.com/NekomyaDev/nudge#readme) for setup instructions, language examples, and an overview of Nudge’s features.

Explore Nudge

• Typed LLM calls — Define the structure you expect from model responses.
• Typed decisions — Express choices, yes/no questions, and scores through a dedicated decision interface.
• Replay and tracing — Inspect agent execution and reproduce recorded runs.
• Budget contracts — Define limits for agent workflows.
• Tools and MCP — Connect agents to external capabilities.
• Parallel workflows — Coordinate work with map, race, and all.

Join the community

Our [Discussions](https://github.com/NekomyaDev/nudge/discussions) are open to conversations about Nudge and the wider world of AI.

Ask for help in Q&A, suggest improvements in Ideas, share your projects in Show and Tell, or talk about AI, agents, models, and coding in General.

Contribute

Found an unclear explanation or have a useful example? Share it in Discussions. Questions and real-world experiences help us make Nudge easier to learn and use.