Web search in n8n AI agents with the SerpCrypto SERP API

Why the HTTP Request node rather than a custom node?

n8n's documentation says the HTTP Request node can run as a regular node or be attached to an AI agent to be used as a tool. That covers a REST API such as SerpCrypto, which takes one GET request with a header and a few query parameters. There is no community node of ours to install or keep up to date, and the credential stays in n8n's encrypted credential store.

The AI Agent node needs a chat model and at least one tool sub-node. Since n8n 1.82.0, every AI Agent works as a Tools Agent, so the agent decides on its own when to call the search tool.

How do I store the API key?

Create a Header Auth credential. n8n's credential documentation describes two fields, the header Name and its Value, and notes that a bearer token is header auth with the name Authorization and the value Bearer <token>. Put Authorization in Name and Bearer plus your SerpCrypto key in Value.

n8n also lists a generic Bearer auth type; either one sends the same header. Give n8n its own API key, so you can see its usage apart from your other scripts and revoke it alone if needed. API keys you can create and revoke at any time from your dashboard.

How does the model fill in the search query?

Let the model define q. Next to an eligible parameter field, n8n shows a star button that hands the value to the model. You can also type the $fromAI() expression yourself, which takes a key, a description, a type and a default value; the description tells the model what to put there. n8n documents that $fromAI() works only with tools connected to the AI Agent node.

Keep gl and hl fixed if your agent serves one market. Let the model set them too if users ask about different countries, and describe the expected codes, such as gb or de.

{{ $fromAI("query", "The search query, as a person would type it", "string") }}

Why turn on Optimize Response?

A full results page in JSON is long, and everything the tool returns goes into the model's context. Optimize Response is a tool-only option of the HTTP Request node that cuts the data passed to the model; n8n notes it can reduce costs. With the JSON response type, set Field Containing Data to organic_results and select only title, link and snippet.

The agent then reads about ten short entries per search instead of every block of the page. If your agent answers local questions, point Field Containing Data at local_results in a second tool instead.

How do I use the same API in a normal workflow?

Without the agent, use the HTTP Request node as a regular node with a fixed q, for example in a Schedule Trigger workflow that checks a keyword every morning. Then add a Split Out node with Field to Split Out set to organic_results: it turns the list into one item per result, ready for a spreadsheet or a database node.

Each run is one search per keyword, so 50 keywords a day use about 1,500 credits a month. The 25,000 searches pack ($18.75) covers that more than ten times over, and Prepaid search credits never expire: unused searches stay on your account until you use them.

How do I test the request outside n8n?

Run the same request in a terminal. If it works there, any problem in n8n comes from the node settings; if it fails, the error body names the cause, as listed on the errors page. Errors are never charged.

# The request the HTTP Request node sends, for a test in a terminal
curl -sG "https://serpcrypto.com/api/v1/search" \
  -H "Authorization: Bearer $SERPCRYPTO_API_KEY" \
  --data-urlencode "q=n8n self-hosted vs cloud" -d gl=us -d hl=en \
  | jq '.organic_results[]? | {title, link, snippet}'