Agent Memory & Multi-Agent Workflows for ComfyUI

20 custom nodes that give ComfyUI agents persistent memory, cross-agent messaging, multi-turn chat, and LLM tool use — all running locally.

● 20 Nodes ● 7 Demo Workflows ● 76 Tests Passing ● 100% Offline

Node Categories

Every node your agents need — identity, memory, messaging, LLM, and utilities.

Identity & Config

Setup & SystemPrompt — give each agent a unique address and personality. Config flows through the graph like a passport.

🔗

Persistent Memory

Remember, Recall, Note — store and search facts that survive across sessions and queue runs.

Multi-Agent Messaging

Send, Inbox, Relay, ChatHistory — agents exchange messages through a file-based mailbox. No server required.

🤖

LLM & Tool Use

LLM, ToolParser — call any OpenAI-compatible endpoint and parse inline tool calls from the response.

🔧

Text Utilities

TextCombine, TextDisplay, PromptTemplate, ConditionalRouter, BatchSplitter — plumbing for any pipeline.

🚀

Zero Dependencies

Pure Python, no pip installs. Works with any local LLM: Ollama, LM Studio, llama.cpp, vLLM — anything with an OpenAI-compatible API.

Example Workflows

Click any card to see the full breakdown — flow diagram, node table, and features.

🎨

Character Design Studio

Memory LLM Tool Use

Consistent character creation for Stable Diffusion. The art director remembers previous characters and maintains visual coherence across prompts.

View Workflow →
📖

Novel Crafter

Memory LLM × 2 Relay

Dual-agent writing pipeline: a writer drafts chapters and sends them to an editor for feedback. Cross-agent messaging via Relay.

View Workflow →
🎬

Story → Video Pipeline

Memory LLM BatchSplitter

Break a story into visual scenes for animation. The director generates scripts + SD prompts, then BatchSplitter separates each scene.

View Workflow →
💬

Multi-Turn Chat

ChatHistory LLM PromptTemplate

Persistent conversation that survives ComfyUI queue resets. History is saved to disk and reloaded next run — true multi-turn chat.

View Workflow →

Agent Debate

Relay Inbox LLM × 2

Two LLM agents debate a topic. Agent A makes an argument and sends it to Agent B, who reads the inbox and writes a rebuttal.

View Workflow →
🌎

World Builder

Memory LLM PromptTemplate

Incremental world creation for games and fiction. Each run adds a location, character, or item — all cross-referenced in memory.

View Workflow →
🔍

RAG Pipeline

Memory ConditionalRouter LLM

Store documents with Remember, retrieve with Recall, answer with LLM, and route errors gracefully with ConditionalRouter.

View Workflow →

Quick Install

Drop into your ComfyUI custom_nodes folder — zero pip installs required.

cd ComfyUI/custom_nodes
git clone https://github.com/gregorkoch/comfyui-agentazall.git
# restart ComfyUI — 20 nodes appear under AgentAZAll/
✓ ComfyUI 0.3.x ✓ Python 3.10+ ✓ Any OpenAI-compatible LLM ✓ Windows / Linux / macOS