AppendAgent
When one agent just isn't enough, jam two together
- Agent1
- Agent2
- Agent
This is the pack's glue node, and it does exactly one thing: takes two agent inputs and gives you a single agent output. In NodeGPT - an early experiment that wired Microsoft's AutoGen multi-agent framework into ComfyUI - agents are objects, not static text, and some downstream nodes expect a list of them rather than one at a time. AppendAgent is how you build that list without leaving the graph.
Why would you? Picture the GroupChat workflow. It takes Agent, Agent2, and up to three optional more. If you're generating a squad programmatically - say an Assistant plus a Critic plus an Engineer, all sharing one LLM - you don't wire three separate agent streams; you chain them into one. AppendAgent is the operator that fuses Agent1 and Agent2 into a combined Agent you can keep appending to. It's the + in your "build the team" step.
Inputs and output
Exactly as thin as it looks:
- Agent1 - first agent (or agent list) to merge.
- Agent2 - second one.
Output: Agent, the merged result. Chain three of these nodes in a row and you've assembled a four-agent roster that you can drop into a GroupChat or any node that eats an Agent.
Honest words about its place in the pack
AppendAgent only makes sense if you're committing to this pack's way of doing things, which has an expiration date stamped on it. The README calls the whole thing "under development, and not everything is functioning correctly yet," and the repo is now gone - as of this writing xXAdonesXx/NodeGPT returns 404 and the author's account is deleted, so the README's install path (cd ComfyUI/custom_nodes && git clone ...) is dead and ComfyUI Manager can't resurrect it. Community reports from when it worked described chats that wouldn't terminate and agents that echoed their input back at each other.
None of that makes the node unteachable. If you load an old NodeGPT workflow that uses it, you now know what it's for: merging agent objects so the group chat gets a full roster. And if you're building today, the lesson carries over - in maintained multi-agent setups you'll find the same "combine" step, just with a less cryptic name.
Installing
cd ComfyUI/custom_nodes
git clone https://github.com/xXAdonesXx/NodeGPT
Restart ComfyUI so it auto-installs requirements (the pack depends on pyautogen). Just know the clone will fail today - you'd be hunting down the fork for the historical early build instead. For a thin utility node like this, the smartest move in 2026 is usually to skip the pack and read what AutoGen's group-chat patterns do, then reach for a maintained pack.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| Agent1 | Agent | — | |
| Agent2 | Agent | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Agent | Agent | — |