Description

n8n-nodes-kong-ai-gateway


An n8n community node package that integrates Kong AI Gateway into your n8n workflows.

It uses the OpenAI-compatible endpoints exposed by Kong’s AI Proxy plugin, allowing you to route requests to any LLM provider (OpenAI, Azure OpenAI, Anthropic, Cohere, Mistral, Llama, and others) through a single, centrally managed gateway.

Nodes

This package exposes three nodes:

• Kong AI Gateway – Chat Model — LangChain-compatible chat model for use inside n8n AI Agent workflows
• Kong AI Gateway – Chat — Standard chat node for direct, non-agent LLM calls
• Kong AI Gateway – Embeddings — Embeddings node for vector generation workflows

—

Installation

In your n8n instance, go to Settings > Community Nodes and install:

n8n-nodes-kong-ai-gateway

Or via npm:

npm install n8n-nodes-kong-ai-gateway

—

Credential Setup

All three nodes share the same credential type: Kong AI Gateway API.

1. In n8n, go to Credentials > New Credential
2. Search for Kong AI Gateway API
3. Fill in the fields:

| Field | Description |
|—|—|
| Base URL | The base URL of your Kong Gateway (e.g. https://kong.example.com) |
| API Key Header Name | The HTTP header used for authentication. Defaults to Authorization. If you use Kong’s Key Auth plugin, set this to apikey. |
| API Key Header Value | The value for the header (e.g. Bearer sk-xxx for Authorization, or your raw API key for Key Auth) |

4. Click Test to verify the connection.

—

Kong-Side Setup

Each node requires a Kong route configured with the AI Proxy plugin.

Example declarative configuration (YAML):

formatversion: "3.0"
services:
  - name: ai-gateway
    url: https://api.openai.com
    routes:
      - name: chat-route
        paths:
          - /llm/chat
        strip_path: true
    plugins:
      - name: ai-proxy
        config:
          route_type: llm/v1/chat
          auth:
            header_name: Authorization
            headervalue: "Bearer OPENAIAPIKEY>"
          model:
            provider: openai
            name: gpt-4o

For embeddings, set route_type: llm/v1/embeddings on the corresponding route.

Refer to the Kong AI Gateway documentation for full configuration options, including rate limiting, request transformation, logging, and multi-provider routing.

—

Node Reference

Kong AI Gateway – Chat Model

A LangChain-compatible chat model designed for use inside the n8n AI Agent node. Connects to Kong’s Chat Completions API or Responses API and supports tool calling.

| Parameter | Default | Description |
|—|—|—|
| Output Format | Completion | Completion uses the Chat Completions API (messages format). Responses uses the Responses API (input format). |
| Route Path | /llm/chat | Kong route path for the Chat Completions API. Shown when Output Format is Completion. |
| Responses Route Path | /responses | Kong route path for the Responses API. Shown when Output Format is Responses. |
| Model (Optional) | (empty) | Model name to include in the request (e.g. gpt-4o). Leave empty if Kong selects the model automatically via the AI Proxy plugin. |
| Max Tokens | 1024 | Maximum number of tokens to generate. |
| Temperature | 1 | Sampling temperature (0–2). |
| Max History Messages | 12 | Maximum number of messages to retain in conversation history before sending to Kong. The system prompt is always preserved and never counted. History trimming happens client-side inside the node — older messages are dropped to prevent Kong from rejecting long multi-turn conversations. Set to 0 to disable trimming. |

Notes:

• Tool calling is supported in Completion mode only.
• In Completion mode, the node correctly handles tool call pairs (assistant + tool result messages) during history trimming — incomplete pairs are never sent to the API.

—

Kong AI Gateway – Chat

A standard chat node for direct LLM calls, without the AI Agent context.

| Parameter | Default | Description |
|—|—|—|
| Output Format | Completion | Completion returns the full Chat Completions response. Responses returns the Responses API format. |
| Route Path | /llm/chat | Kong route path for the Chat Completions endpoint. |
| Model (Optional) | (empty) | Model name to include in the request. Leave empty if Kong manages model selection. |
| Max Tokens | 4096 | Maximum number of tokens to generate. |
| Temperature | 1 | Sampling temperature (0–2). |
| Max History Messages | 30 | Maximum number of messages to retain in conversation history. Same trimming logic as the Chat Model node. |

—

Kong AI Gateway – Embeddings

Generates vector embeddings via Kong’s embeddings endpoint.

| Parameter | Default | Description |
|—|—|—|
| Route Path | /llm/embeddings | Kong route path for the embeddings endpoint. |
| Model (Optional) | (empty) | Model name to include in the request. Leave empty if Kong manages model selection. |
| Dimensions | 512 | Number of dimensions for the output embedding vectors. |
| Batch Size | 512 | Number of inputs to process per request. |

—

Known Limitations

• No streaming support — Nodes wait for the complete response before returning. SSE streaming is not supported.
• No function/tool calling in Chat node — Tool use is only available in the Chat Model node when used inside an AI Agent.
• Responses API does not support tool messages — Tool calling is only available in Completion mode.

Install dependencies
npm install
Build
npm run build
Watch mode
npm run dev
Lint
npm run lint