n8n

n8n Integration

Use all 28 CrawlForge web scraping tools inside n8n workflows. Connect natively through n8n's MCP Client node over Streamable HTTP, or call the REST API from the HTTP Request node — both paths are verified end-to-end.

Two Ways to Connect

n8n can talk to CrawlForge over two protocols. Pick the one that matches where your n8n runs.

MCP Client Node

Native Model Context Protocol connection over Streamable HTTP. n8n discovers all 28 tools automatically and renders each tool's parameters as form fields. Requires a CrawlForge MCP server endpoint you host (one npx command).

Native MCP

HTTP Request Node

Plain REST calls to crawlforge.dev — every tool is a POST endpoint. Nothing to host, works on n8n Cloud and self-hosted alike.

REST API

Option 1: MCP Client Node

n8n ships a built-in MCP Client node (plus an MCP Client Tool sub-node for AI Agent workflows). It speaks MCP's Streamable HTTP transport — the same transport the CrawlForge MCP server exposes in HTTP mode. This guide was verified end-to-end on n8n 2.1.3: the node connects, lists all 28 tools, and executes them.

Start the CrawlForge MCP server in HTTP mode

The npm package runs over stdio by default; the --http flag switches it to Streamable HTTP. Your API key does double duty: it authenticates the server to CrawlForge, and it is the Bearer token n8n must present.

Bash

CRAWLFORGE_API_KEY=cf_live_your_api_key_here \
npx crawlforge-mcp-server --http

# MCP endpoint:  http://localhost:10000/mcp
# Health check:  http://localhost:10000/health
# Set PORT to listen somewhere other than 10000

Add the MCP Client node and connect

Server Transport: HTTP Streamable (the default)
MCP Endpoint URL: http://localhost:10000/mcp — or http://host.docker.internal:10000/mcp when n8n runs in Docker on the same machine
Authentication: Bearer Auth — create a credential whose token is the same CrawlForge API key the server was started with

Pick a tool and execute

Open the Tool dropdown — n8n connects to the server and lists all 28 CrawlForge tools. Select one (e.g. fetch_url), and with Input Mode: Manual n8n renders the tool's parameters as form fields. Click Execute step to run it; the tool result appears as the node's output, ready to wire into any downstream node.

n8n in Docker? A containerized n8n cannot reach localhost on your host machine. Use http://host.docker.internal:10000/mcp as the endpoint URL instead.

AI Agent workflows: The same connection works from the MCP Client Tool sub-node attached to an AI Agent node — your agent then calls CrawlForge tools autonomously. Use Tools to Include to limit which of the 28 tools the agent may call and keep credit spend predictable.

Option 2: HTTP Request Node

Every CrawlForge tool is also a REST endpoint at https://www.crawlforge.dev/api/v1/tools/<tool_name>. This needs no server of your own, so it is the right choice for n8n Cloud (which cannot reach a localhost MCP endpoint). Add an HTTP Request node and configure it:

HTTP Request nodeJson

// Method: POST
// URL: https://www.crawlforge.dev/api/v1/tools/extract_content

// Headers
{
  "X-API-Key": "cf_live_your_api_key_here",
  "Content-Type": "application/json"
}

// Body (JSON)
{
  "url": "https://example.com/article"
}

The n8n workflow automation guide builds a complete production pipeline on this pattern — reusable credentials, a scheduled price monitor, error handling, and Slack notifications. Tool parameters and response shapes are in the API reference.

Which option should you use?

	MCP Client node	HTTP Request node
Works on n8n Cloud	Only with a publicly hosted MCP server	Yes
Runs your own server	Yes — one `npx` command	No
Tool discovery	Automatic — all 28 tools listed in the node	Manual — one endpoint per tool
Parameter input	Generated from the tool's schema	Hand-written JSON body
AI Agent tool use	Yes, via MCP Client Tool sub-node	No

Hosting the MCP endpoint publicly: To use the MCP Client node from n8n Cloud, host the server on a public URL — the crawlforge-mcp-server repo ships a Dockerfile and a Render blueprint that run it in HTTP mode.

Credit Costs

Both connection methods bill the same per-tool credit costs against your API key — for example fetch_url is 1 credit, extract_content is 2, and search_web is 5. The full table is in the API reference; the free plan's one-time 1,000 credits are plenty for building and testing workflows.

Troubleshooting

Symptom	Fix
401 Unauthorized from the MCP Client node	The Bearer credential must exactly match the `CRAWLFORGE_API_KEY` the server was started with — they are compared as strings.
Connection refused / could not connect	The server isn't running in HTTP mode, or n8n runs in Docker and the endpoint says `localhost` — use `host.docker.internal`.
Tool dropdown stays empty	Check Server Transport is HTTP Streamable and the endpoint path ends in `/mcp`. Verify the server is healthy: `curl http://localhost:10000/health`.
402 Payment Required on tool execution	The account behind the API key is out of credits — check usage on the dashboard.

Start Automating with n8n

Sign up free and get 1,000 credits — no credit card required. Then follow the workflow guide to build your first scraping pipeline.

Get Your Free API Key n8n Workflow Guide

CRAWLFORGE_API_KEY=cf_live_your_api_key_here \ npx crawlforge-mcp-server --http # MCP endpoint: http://localhost:10000/mcp # Health check: http://localhost:10000/health # Set PORT to listen somewhere other than 10000

// Method: POST // URL: https://www.crawlforge.dev/api/v1/tools/extract_content // Headers { "X-API-Key": "cf_live_your_api_key_here", "Content-Type": "application/json" } // Body (JSON) { "url": "https://example.com/article" }

MCP Client node

HTTP Request node

Works on n8n Cloud

Only with a publicly hosted MCP server

Yes

Runs your own server

Yes — one npx command

Tool discovery

Automatic — all 28 tools listed in the node

Manual — one endpoint per tool

Parameter input

Generated from the tool's schema

Hand-written JSON body

AI Agent tool use

Yes, via MCP Client Tool sub-node

Symptom

Fix

401 Unauthorized from the MCP Client node

The Bearer credential must exactly match the CRAWLFORGE_API_KEY the server was started with — they are compared as strings.

Connection refused / could not connect

The server isn't running in HTTP mode, or n8n runs in Docker and the endpoint says localhost — use host.docker.internal.

Tool dropdown stays empty

Check Server Transport is *HTTP Streamable* and the endpoint path ends in /mcp. Verify the server is healthy: curl http://localhost:10000/health.

402 Payment Required on tool execution

The account behind the API key is out of credits — check usage on the dashboard.