🧠 Conversation Memory

Store, retrieve, and clear multi-turn conversation history for AI workflows.

Overview

The Conversation Memory node enables multi-turn conversations in your workflows by persisting chat history across executions. Store user and assistant messages, retrieve recent context for AI prompts, and auto-summarize older messages to keep token usage efficient.

Actions

  • Retrieve — fetch recent messages and an optional summary for a conversation. Output includes messages array, summary text, and total message count.
  • Store — save a message with a role (user, assistant, or system) and content. Supports variable interpolation for dynamic content.
  • Clear — permanently delete all messages and summaries for a conversation. Use when a conversation should be reset.

Configuration

  • Action — retrieve, store, or clear
  • Conversation ID — a unique identifier for the conversation (supports {{variables}}, e.g. {{trigger.from}} for SMS threads)
  • Recent Messages — how many recent messages to retrieve (default: 10)
  • Include Summary — whether to include auto-generated summaries of older messages
  • Output Variable — variable name to store retrieved memory (default: memory)

AI Prompt Integration

The AI Prompt node has built-in memory support. Enable "Conversation memory" in the AI Prompt config to automatically load history, send it with your prompt, and store both the user message and assistant response — without needing separate Memory nodes.

Common Use Cases

  • SMS chatbot — use {{trigger.from}} as the conversation ID so each phone number maintains its own thread
  • Support agent — persist context across webhook calls so the AI remembers prior questions and answers
  • Long-running conversations — enable auto-summarization to compress older messages and stay within LLM context limits