• ACCURACY
  • CONTENT
  • REPAIR
  • HUMAN

How to debug RAG retrieval with your AI client

Connect R2R to your AI client once. From there, you can audit your retrieval logic and inspect the exact data feeding your context window.

Ask AI about this page

Short answer

How do I debug RAG retrieval with my AI?

Your AI uses R2R tools to search your index and pull specific document chunks. This lets you see exactly what the retriever found and why it chose those pieces of data. You'll see the specific outcomes of these checks below.

Debugging outcomes.

Where your findings land.

What you get after running these checks.

ACCURACY

Verify search relevance

The AI runs searches against your R2R index. You'll see if the semantic results actually match your intent.

CONTENT

Inspect raw chunks

Your AI pulls the specific text segments used in context. You can check if the chunking strategy is working.

REPAIR

Identify retrieval gaps

Spot where the retriever fails to find relevant data. This helps you decide if you need better embeddings or better chunking.

HUMAN

Manual index tuning

The AI cannot rewrite your R2R index or change your embedding model. You'll need to handle those changes yourself.

The workflow

What your AI does when the request arrives.

The AI acts as an auditor for your R2R instance.

  1. Query the index

    Your AI sends a query to R2R to see what the semantic engine returns.

    search

  2. Fetch specific segments

    Once a relevant chunk is identified, the AI pulls the full content of that specific piece.

    get_document

  3. Compare results

    The AI uses the retrieved text to evaluate if the search was actually successful.

    searchget_document

  4. Report findings

    The AI presents the retrieved text and its assessment of the retrieval quality.

    get_document

Try it

Copy these to start.

Use these prompts to begin your debugging session.

Starting points

Swap out the bracketed text with your specific search terms or document IDs.

R2R Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_Psy9XiP4IhGY7SchHirciLtxL89aKlE6LAoJStd2/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — R2R capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "r2r-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_Psy9XiP4IhGY7SchHirciLtxL89aKlE6LAoJStd2/mcp"
    }
  }
}
Copy into chat04

  • Search the R2R index for 'quarterly revenue report 2023' and tell me if the results are relevant.

  • Find the document with ID 'doc_992' and show me the text in the first chunk.

  • Run a search for 'employee benefits policy' and check if the retrieved chunks contain the section on dental insurance.

  • Search for 'project phoenix timeline' and tell me if any of the retrieved documents are outdated.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect R2R once, then ask.

Just link your R2R instance to Vinkius. Your credentials stay encrypted and you can immediately start debugging with any MCP-compatible client.

Connect R2R to your AI

FAQ

How this task behaves.

  • 01

    Can the AI change my R2R embedding model?

    No. The AI can only read from the index using the existing tools.

  • 02

    Can the AI delete documents from my R2R instance?

    No. This MCP only provides tools to search and get document content.

  • 03

    How does the AI know if a search result is good?

    The AI compares your original question against the text it pulls via the search and get_document tools.

  • 04

    Can I use this with Cursor?

    Yes, any MCP-compatible client like Cursor, Claude, or Windsurf works.

  • 05

    Does the AI see my whole database at once?

    No. The AI only sees the specific chunks it retrieves through the search tool.

  • More questions about R2R? See everything the R2R Connector can do

Connect R2R to Claude, Cursor, ChatGPT & more