RSS Tool

The agent-callable sibling of this pack's RSS Loader. That node fetches one fixed feed URL every run, set by you on the node. This one has no URL field at all - just an enable toggle - because the feed to check isn't decided ahead of time; the LLM supplies it itself at call time, deciding both whether checking a feed is the right move for the conversation and which one to check.

How it works

Wire the tool output into an LLM/agent node's tool-list input. When the model decides fetching a feed would help - answering a question about current news, checking for updates on something specific - it calls this tool with a feed URL of its own choosing, and the result flows back into the conversation.

The inputs and outputs that matter

• is_enable - the only input, the pack's standard bypass toggle.

One output: tool (STRING) - wire it into your agent's tools input.

How to install it

• ComfyUI Manager: search "comfyui_LLM_party", install, restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then pip install -r requirements.txt from inside the pack folder using ComfyUI's own Python, then restart.

No API key of its own beyond whatever LLM node you've already configured - it's the model that decides to use it, not this node acting independently.

Where it fits

Reach for RSS_loader when you already know which feed you want checked every run - a fixed news source, a specific blog, something you're monitoring on a schedule. Reach for this tool version when the feed itself should be part of the agent's judgment - a research assistant that decides which outlet to check based on what it's being asked, or a general-purpose agent you don't want to pre-wire with one specific source.

Common issues & troubleshooting

The LLM never uses the tool. Same as every tool-calling node in this pack - whether the model reaches for it depends on its persona or system prompt establishing that it can. If nothing in the agent's instructions mentions checking feeds, it may just answer from what it already knows instead.

The call fails or comes back empty. The agent-supplied feed URL has to actually be a valid, reachable RSS/Atom feed - a model can pick a URL that looks plausible but doesn't resolve, or points at a webpage rather than an actual feed, the same failure mode as manually typing a bad URL into RSS Loader.

The model hallucinates feed results instead of real ones. Confirm the LLM node this tool is wired into genuinely supports tool/function calling - a model or API configuration that doesn't support it will sometimes generate a plausible-sounding "result" instead of actually invoking the tool at all.