Skip to content

New AI assistant answers: ChatGPT, Gemini, Copilot and AI Mode by API

SerpKite
Get API key

No-code · n8n

n8n + SerpKite

Install the SerpKite community node and search Google, read and crawl websites, get cited answers and ask AI assistants from any n8n workflow or AI Agent. One API key credential, no code.

The SerpKite node is on npm

Install n8n-nodes-serpkite on self-hosted n8n. n8n Cloud offers community nodes once n8n has verified them, which is pending; on Cloud, use the HTTP Request setup below for now.

Overview

What n8n is, and where SerpKite fits

n8n is a workflow automation tool with a strong AI agent story. The SerpKite community node (n8n-nodes-serpkite on npm, published with provenance) adds SerpKite as a regular node: pick a resource and an operation, fill in the fields, and every operation also works as a tool for the AI Agent node.

It covers search across Google's verticals, Maps and reviews, webpages and whole sites, cited answers and research reports, and the answers of ChatGPT, Gemini, Copilot Search, Google AI Mode and Brave. Each operation offers exactly the options its endpoint accepts.

Anything else, such as batches or monitors, is one built-in HTTP Request node away, shown below with an importable workflow.

Setup

Set it up in 5 steps

  1. 1

    Get an API key

    Sign up (no card), create a key in the dashboard and export it as SERPKITE_API_KEY. New accounts get 2,500 free credits, then 1,000 every month.

    export SERPKITE_API_KEY=skt_live_…
  2. 2

    Install the community node

    In self-hosted n8n go to Settings → Community nodes → Install, enter n8n-nodes-serpkite and confirm.

  3. 3

    Add your API key

    Add a SerpKite node to a workflow. Under Credential choose Create new credential and paste your full skt_live_… key. Testing the credential spends no credits.

  4. 4

    Pick a resource and an operation

    For example Search → Web Search with the query {{ $json.query }}, or AI Assistant → Ask ChatGPT with a prompt. Options lists every parameter the operation takes.

  5. 5

    Use the results

    Each input item returns one response with results and meta. Add a Split Out node on results to get one item per result.

Code

Importable workflows

Copy a workflow and paste it into an n8n canvas (Ctrl/Cmd+V), then pick your credential. The first needs the community node; the second uses only core n8n and a Header Auth credential (Authorization: Bearer skt_live_…).

{
  "name": "SerpKite Google search",
  "nodes": [
    {
      "parameters": {},
      "name": "When clicking 'Execute workflow'",
      "type": "n8n-nodes-base.manualTrigger",
      "typeVersion": 1,
      "position": [0, 0]
    },
    {
      "parameters": {
        "resource": "search",
        "operation": "search",
        "q": "best crm for startups",
        "options": { "country": "us", "num": 10 }
      },
      "name": "SerpKite",
      "type": "n8n-nodes-serpkite.serpKite",
      "typeVersion": 2,
      "position": [220, 0]
    }
  ],
  "connections": {
    "When clicking 'Execute workflow'": {
      "main": [[{ "node": "SerpKite", "type": "main", "index": 0 }]]
    }
  }
}

Operations

What the SerpKite node does

Search: Web Search (organic results, answer box, knowledge graph, People Also Ask; optionally the top pages as Markdown and a cited AI answer), News Search, Image Search, Video Search, Shopping Search, Scholar Search, Patent Search and Autocomplete.

Place: Maps Search and Local Search (places with address, coordinates, rating and hours), Get Place (the full listing of one place) and Get Reviews.

Web: Fetch Webpage (a page or PDF as Markdown), Extract Pages (up to 20 URLs with query-ranked highlights, an AI summary or JSON for a schema), Map Site (a site's URLs) and Start Crawl / Get Crawl (a whole site or section as Markdown).

AI Research: Answer (a short answer citing live Google results) and Start Research / Get Research (a multi-step research report with sources).

AI Assistant: Ask ChatGPT, Ask Gemini, Ask Copilot, Ask AI Mode and Ask Brave AI, each returning the answer with its cited sources and products. Account: Get returns your balance, limits and usage.

Crawls and research run in the background: turn on Wait for Completion to get the finished result in the same step, or start them in one step and fetch them later by ID. Every operation is billed per the pricing page; failed and empty requests are free.

AI Agent node

Give an n8n AI Agent Google search

Connect the SerpKite node to the AI Agent's Tool input and pick an operation, such as Search → Web Search. Let the model fill the query with {{ $fromAI('query', 'what to search Google for') }}, and set Options → Response Format to Markdown or Compact so the agent reads compact text instead of a large JSON object.

Without the node (for example on n8n Cloud), attach an HTTP Request Tool with the URL and Header Auth credential from the second workflow above and this JSON body.

JSON body
{
  "q": "{{ $fromAI('query', 'what to search Google for') }}",
  "country": "us",
  "format": "markdown"
}

Recipes

Workflows people build

Local market map: Schedule → SerpKite Place → Maps Search for plumbers with Location Leeds, England, United Kingdom → Split Out results → Google Sheets.

News monitor: Schedule (hourly) → Search → News Search with Time Window Hour → Filter new links → Slack.

Docs to knowledge base: Web → Start Crawl with Wait for Completion → Split Out result.pages → your vector store.

Rank tracking: Schedule (daily) → HTTP Request to /v1/rank with q and domain → the response's position → Sheets.

AI assistants

Ask ChatGPT, Gemini and other AI assistants

With the SerpKite node, choose AI Assistant → Ask ChatGPT (or Ask Gemini, Ask Copilot, Ask AI Mode, Ask Brave AI), enter the prompt and add Track Brands or Domains under Options. The node waits up to two minutes for the answer and returns it (text, markdown) with cited sources, products and mentions of what you track. As an AI Agent tool, set Response Format to Markdown.

Every assistant is one POST: https://api.serpkite.com/v1/chatgpt (or /v1/gemini, /v1/copilot, /v1/ai_mode, /v1/brave_ai) with the same Bearer header and a JSON body with prompt. In an HTTP Request (or HTTP Request Tool) node instead, raise Options → Timeout to 120000 ms, and add "format": "markdown" when an AI Agent reads the answer. The response has the answer (text, markdown), cited sources, products and, with track, mentions of your brands and domains. See AI assistants.

To track answers over time without a long-running step, create a monitor once (endpoint: "chatgpt", prompt, track, webhook_url): every run is POSTed to your webhook with the answer, its new_sources and mentions.

ChatGPT 4, Gemini 3, Copilot Search 3, Google AI Mode 2 and Brave AI 2 credits per answer; failed, refused and empty answers are free. Brave answers in 1–10 s, ChatGPT in 10–30 s and Gemini in 5–30 s, so allow at least 120 seconds per call.

JSON body
{
  "prompt": "What are the best CRMs for a 10-person startup?",
  "track": ["hubspot.com", "Pipedrive"],
  "country": "us"
}

FAQ

n8n and SerpKite: common questions

Start building

Give your n8n project Google search

2,500 free credits, then 1,000 every month. No credit card.