n8n-nodes-k2think

An n8n community node for K2 Think V2 (MBZUAI-IFM) — the reasoning model served from the OpenAI-compatible gateway at api.k2think.ai.

This is an action node: drop it into any workflow to send a prompt to K2 Think and get the answer back. It handles the gateway's quirks for you:

• uses max_completion_tokens (not max_tokens)
• non-streaming
• automatically strips the <think>…</think> reasoning preamble (toggleable)
• adds token headroom so the reasoning never truncates the answer

Note on the AI Agent node: n8n's "Chat Model" sub-nodes (the kind that plug into the AI Agent / Chains) are built on LangChain, which requires a runtime dependency and therefore cannot be verified for n8n Cloud. This package is a dependency-free action node so it can be verified and installed on Cloud. To use K2 Think as an Agent model on Cloud today, use the built-in OpenAI Chat Model node with a credential whose Base URL is https://api.k2think.ai/v1.

Installation

n8n Cloud / Self-hosted UI

Settings → Community Nodes → Install → n8n-nodes-k2think (available after npm publish + n8n verification).

Self-hosted (manual)

cd ~/.n8n/nodes
npm install n8n-nodes-k2think

Credentials

Create a K2 Think API credential:

• API Key — your api.k2think.ai key
• Base URL — defaults to https://api.k2think.ai/v1

Usage

Add the K2 Think node, pick the credential, set:

• Model — MBZUAI-IFM/K2-Think-v2
• Prompt — your user message
• System Prompt (optional)
• Options — temperature, max completion tokens, reasoning headroom, strip <think>, output field, include raw

License

MIT