Rig is an open-source Rust library for building modular, scalable LLM applications. It gives Rust developers ergonomic abstractions for the things LLM apps all need — model providers, agents, embeddings, vector stores and retrieval pipelines — without leaving the Rust ecosystem. The project has earned about 8,900 stars on GitHub and is actively maintained in 2026.
Rig provides a unified interface over multiple model providers, so the same application code can talk to different LLMs. On top of that it offers agent abstractions with tool use, completion and embedding APIs, and integrations with vector stores for retrieval-augmented generation. Because it is a library rather than a hosted service, you compose the pieces you need and keep full control of your application’s architecture, which suits teams building latency-sensitive or embedded AI features in Rust. Documentation and examples live on the project site at rig.rs. Rig’s ecosystem includes integrations with a wide range of model providers — OpenAI, Anthropic, Cohere, Google, Mistral, Ollama and others — and with vector databases such as MongoDB, LanceDB, Neo4j, Qdrant and SurrealDB. Agents built with Rig can be given tools, long-term memory through those stores, and multi-turn conversation handling, and the whole stack compiles into your own binary, which keeps deployment simple and avoids a separate Python service in Rust-first stacks.
Rig is completely free and open source under the MIT license. You only pay for the model APIs or infrastructure your application uses.
Rig is for Rust developers who want to build production LLM applications — agents, RAG pipelines and AI-backed services — with a typed, modular toolkit instead of raw HTTP calls.