Why is my n8n RAG Agent failing to receive vectorstore responses, causing max iteration errors?

Below are a few things to check and adjust since the error “Agent stopped due to max iterations” usually means that even though the tool (in this case your vectorstore tool) returns some data, the agent never receives what it expects to finish its loop. Here are some troubleshooting steps:

1. Validate the tool’s output field:
   Make sure that the vectorstore tool is configured to send back a valid, non‑null string to the agent. The AI Agent expects the tool’s response in a particular field (often something like chatInput or content), so verify that the mapping is correct and that the output isn’t null.

2. Check prompt settings:
   If the agent is set to “Take from previous node automatically” but the incoming data contains null or isn’t matching the expected field, the agent might loop endlessly. Try switching the Prompt setting to “Define below” so you can manually reference the specific field from the vectorstore tool.

3. Ensure sub-node compatibility:
   If you’re using any memory or tool sub‐nodes (like the Window Buffer Memory node), ensure that you’re running the latest version. Sometimes older versions may not pass data along as expected. If in doubt, remove and re‑add the node to guarantee you’re using the current version.

4. Confirm the Chat Model connection:
   Although the error isn’t exactly “A Chat Model sub-node must be connected,” it can be related if the agent never properly hands off to a Chat Model or process that terminates its iterative loop. Double‑check that your AI Agent has the appropriate Chat Model sub‑node connected, as required in many workflows.

Following these steps should help the agent complete its process rather than hitting the maximum iteration limit. If the problem persists after checking these points, consider simplifying the workflow to isolate where the response is lost or reaching out on the n8n community forum with details of your node setup.