Rig

Rig is an open-source Rust library for building LLM-powered applications and AI agents with a focus on performance, type safety, and composability. For teams building high-performance agent systems where latency, memory safety, and reliability are critical, Rig provides a Rust-native alternative to Python-based frameworks like LangChain and CrewAI.

The framework provides high-level abstractions for LLM completion and embedding workflows through its provider-agnostic API. Rig supports OpenAI, Anthropic (Claude), Google Gemini, Cohere, and Perplexity as LLM providers, with a unified interface that makes switching between providers trivial. The library handles prompt engineering, context management, and response parsing with Rust's type system ensuring correctness at compile time.

Rig's RAG (Retrieval Augmented Generation) support includes vector store integrations with MongoDB, LanceDB, Neo4j, and Qdrant, with an extensible trait system for custom vector store backends. The pipeline system allows composable, reusable agent workflows where each stage is a typed transformation, catching integration errors before runtime.

For agent tool use, Rig supports structured function calling with automatic schema generation from Rust types, making it easy to give agents typed tools that are guaranteed to receive valid inputs. The framework is built on Tokio for async runtime, making it excellent for high-concurrency agent servers.

Rig is gaining traction among teams building performance-critical agent infrastructure, API servers that handle thousands of concurrent agent requests, and embedded systems where Python's overhead is unacceptable. The Rust ecosystem's package manager (Cargo) makes Rig easy to integrate into existing Rust projects.