Alternatives to OpenAI Cookbook

Guidance is an open-source Python library for constrained LLM generation with regex, context-free grammars, and inline control flow.

MIT-licensed Swift package for calling the OpenAI API with native async/await and Combine on Apple platforms.

Mellea turns type-annotated Python functions into validated, testable LLM calls with token-level constrained decoding.

xsAI is an extra-small, OpenAI-compatible AI SDK for JavaScript and TypeScript — 13 KB gzipped with every package included.

Open-source TypeScript AI SDK for streaming AI apps and agents across 100+ models with built-in fallbacks

Outlines enforces structured LLM output during decoding, so JSON, regex, and grammar-constrained generations come back valid on the first

Go SDK for AI apps: one type-safe API for text generation, streaming, tools and MCP across 25+ LLM providers.

A headless React hooks library that runs Vicuna-13B chat entirely in your browser via WebGPU, with no prompt or completion leaving the client.

Open-source AI SDK and toolkit for building agentic apps and agents in TypeScript, Go, Python, and Dart.

Open source Python and TypeScript SDKs for building production AI agent harnesses you control end to end.

Python agent framework for small language models — local engines, OpenAI-compatible servers, or hosted providers, with tool-call tracing

Open Harness is an MIT-licensed TypeScript SDK for building Claude Code-style agents with stateless primitives and composable middleware.

Free browser markdown editor with Battle Mode for comparing OpenAI, Claude, and Gemini prompt outputs, client-side only.

Marvin turns ordinary Python functions, classes, and Pydantic models into LLM-powered AI tasks with decorators like @ai_fn and

Mirascope is an open-source Python library that turns LLM calls, tools, and prompt versioning into plain decorated functions.

BAML is a type-safe, Turing-complete language for agentic and LLM code that compiles typed Python and TypeScript SDKs.

Pydantic AI is the typed Python agent SDK: declare a model string and a Pydantic output type, get validated structured results on every run.

Eino is an open-source Go LLM framework from ByteDance for building agents, graphs, and RAG pipelines in your existing Go backend.

Rig is a Rust LLM framework for building type-safe AI agents across 20+ model providers with one unified API.

Open-source Java library that tokenizes text with OpenAI's BPE encodings so you can count, encode and decode locally

Open-source Rust library for building streaming, parallel LLM pipelines — data indexing, retrieval-augmented generation, and autonomous agents.

Fenic turns messy text into typed, queryable Semantic DataFrames with LLM-powered extraction, classification, and semantic joins in Python.

Open-source (MIT) MCP testing and full-stack MCP App framework for building ChatGPT Apps and Claude Connectors from one React codebase.

MIT-licensed Ruby gem that plugs MCP servers into RubyLLM chat workflows with idiomatic Ruby APIs

Open Strawberry is a free, open-source chain-of-thought inspector for reading, diffing, and comparing model reasoning traces across six LLM

Cross-provider prompt IDE for building, testing, and scoring prompts against 150+ LLMs from 15 providers.

Open-source Ruby gem giving one unified API across Anthropic, DeepSeek, Google, Mistral, Llama, and OpenAI

Open-source Python library for loading, fine-tuning, and running transformer models across text, vision, audio, and video.

An interactive 470-page ebook with runnable Colab notebooks that takes you from LLM theory to production apps.

flompt is a free visual prompt builder that compiles prompts into Claude-optimized XML blocks