Claude/Project/RAG

• Claude Project RAG

  • Retrieval augmented generation (RAG) is how a Claude project holds more knowledge than fits in the model's context window.

  • Notes from the official docs

    • Retrieval augmented generation (RAG) for projects | Claude Help Center
      • RAG for projects is available on paid plans (Pro, Max, Team, and Enterprise).
      • Previously, projects had a knowledge capacity limit based on the context window. Once you reached this threshold, it wasn't possible to add more content. With RAG, you can continue adding knowledge beyond these limits while maintaining full functionality.
      • RAG activates automatically when needed.
      • All existing projects automatically benefit from RAG when the project knowledge exceeds context limits.
    • The same article says RAG can expand capacity by up to 10x, shows a visual indicator when a project is RAG-enabled, and can switch a project back to context-based processing if knowledge drops below the threshold. In RAG mode Claude uses a project knowledge search tool to retrieve relevant passages.

  • Open

    • The point at which RAG starts, and how the capacity percentage behaves once it does, are not stated in the docs. See Claude/Project/Q/What is the limit on the capacity for adding knowledge to a Claude project?.