Chat agent that runs any model through OpenRouter

Chat agent that runs any model through OpenRouter

An n8n chat agent whose language model is a setting. You type a message in the chat, and the agent answers with whichever model
you named in one node. The model call goes through OpenRouter, which fronts models from many companies, so you can switch from
one maker's model to another by changing a single line of text. The agent keeps the last few messages of each chat.

What is inside

• workflow.json: the workflow as n8n exports it, eight nodes. Five of them run and three are sticky notes. It holds no
  credential.
• README.md: these notes.

What we changed from the original copy

• The default model in Settings was deepseek/deepseek-r1-distill-llama-8b. OpenRouter's public model list, read on 12 November
  2026, no longer holds it. The default is deepseek/deepseek-v4-flash now, which the list does hold.
• The sticky note of model examples named five models that the same list no longer holds. It names six that it does: openai/gpt-6-luna,
  google/gemini-3.8-flash, deepseek/deepseek-v4-flash, mistralai/mistral-small-2603, qwen/qwen3.8-flash and
  google/gemma-4-26b-a4b-it:free.
• The credential link and the chat webhook ID are gone, and the file's name is ours.
• Nothing else. The agent, the memory and the model node are as the file had them.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenRouter account and API key.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a credential of the type OpenAI API. Put your OpenRouter key in the key field, and set the base URL to
   https://openrouter.ai/api/v1. Name it something like OpenRouter.
3. Select that credential in LLM Model.
4. Open Settings and check the model value. It is an OpenRouter model ID, such as deepseek/deepseek-v4-flash.
5. Press the Chat button and send a message.

How a run goes

1. When chat message received starts a run for each message and passes on the text as chatInput, with a sessionId.
2. Settings sets three fields: model, which holds the model ID, prompt, which copies the message, and sessionId, which
   copies the chat's session ID.
3. AI Agent (version 1.7) takes {{ $json.prompt }} as its prompt, with the prompt type set to define.
4. LLM Model is the OpenAI Chat Model node, version 1.1. Its model field reads {{ $json.model }} from Settings, and its
   credential points at OpenRouter, so the call goes there.
5. Chat Memory is a Simple Memory node, version 1.3. It uses the custom key {{ $json.sessionId }}, so each chat keeps its own
   history. Its context window length is left at the default, 5.

Change the look

• The model: the model value in Settings, or the model field of LLM Model directly.
• How much the chat remembers: Context Window Length in Chat Memory.
• The agent's behaviour: add a system message in AI Agent, under Options.
• Tools: connect them to AI Agent. The model you pick must support tool calls.
• A chat page for other people: switch on Make Chat Publicly Available in When chat message received.

Good to know

• Model IDs change often. If a model is withdrawn, OpenRouter refuses the call. Check the list at https://openrouter.ai/models
  before you rely on an ID.
• Newer n8n versions have a dedicated OpenRouter Chat Model node, as n8n's source showed on 12 November 2026. This copy uses the
  OpenAI Chat Model node with an OpenRouter address, which also runs on older versions.
• The memory is kept in n8n's own process. It is gone after a restart, and it does not suit a setup that runs several workers.
• A public chat lets anyone who finds its address spend your OpenRouter credit.
• IDs that end in :free carry no price in the model list. They change often, so check one before you rely on it.
• Where the data goes: each message and the chat history go to OpenRouter, which passes them on to the company that runs the model
  you chose.
• Billing: OpenRouter charges by the price of the model you choose, for the tokens of each message and of the history sent with it.