Crawl4AI Plus for n8n

Enhanced fork targeting Crawl4AI v0.8.0 with a progressive-disclosure two-node architecture: a Simple node (4 operations) for general users and an Advanced node (15 operations) for power users.

What's New in v5.2.x

v5.2.1 — README & documentation update

• Documented the new Custom (OpenAI-Compatible) provider setup (see below)

v5.2.0 — Fix broken custom LLM providers

• Replaced broken opencode and LiteLLM/Custom provider options with a single Custom (OpenAI-Compatible) option that works with any OpenAI-compatible API
• Fixed OpenRouter base URL bug (was incorrectly using Ollama's URL)
• Works with: LLM7, OpenCode, Together AI, Fireworks, self-hosted models, or any OpenAI-compatible endpoint

How to set up a Custom provider

1. Set LLM Provider → Custom (OpenAI-Compatible)
2. Enter your Base URL (e.g. https://api.llm7.ai/v1)
3. Enter your Model Name (e.g. mistral-7b)
4. Enter your API Key (optional)

Why this works: Crawl4AI's Docker uses litellm internally. litellm rejects unknown provider prefixes (opencode/, llm7/), but its openai/ prefix + custom base_url routes requests to any compatible endpoint. See CHANGELOG.md for full details.

Project History & Attribution

This is a maintained fork with enhanced features for Crawl4AI 0.8.0.

Fork Chain

• Original author: Heictor Hsiao - golfamigo/n8n-nodes-crawl4j
• First maintainer: Matias Lopez - qmatiaslopez/n8n-nodes-crawl4j
• Current maintainer: Max Soukhomlinov - msoukhomlinov/n8n-nodes-crawl4ai-plus

All credit for the original implementation goes to Heictor Hsiao and Matias Lopez.

v5.0.0 is a breaking change — the 3-node architecture (BasicCrawler, ContentExtractor, SmartExtract) has been replaced with 2 new nodes. Existing workflows will need rebuilding. See CHANGELOG.md for full details.

Features

Crawl4AI Plus — Simple Node (4 operations)

Designed for general users with smart defaults and minimal configuration:

• Get Page Content — Crawl a URL and get markdown (single page, follow links, or full site via crawl scope)
• Ask Question — Ask a question about a page using LLM extraction
• Extract Data — Extract contact info, financial data, or custom structured data (regex presets + LLM)
• CSS Extractor — Extract structured data using CSS selectors

Crawl4AI Plus Advanced — Advanced Node (15 operations in 3 groups)

Full API control via 3 standardized collections (Browser & Session, Crawl Settings, Output & Filtering):

Crawling

• Crawl URL — Single URL with full browser/crawler/output configuration
• Crawl Multiple URLs — Manual list or recursive discovery (BFS/DFS/BestFirst strategies)
• Stream Crawl — Streaming via /crawl/stream for large URL sets
• Process Raw HTML — Process pre-fetched HTML without a network request
• Discover Links — Extract, filter, and score links (internal/external, include/exclude patterns)

Extraction

• LLM Extractor — AI-powered structured extraction with schema support
• CSS Extractor — Structured extraction using JsonCssExtractionStrategy
• JSON Extractor — Extract JSON from direct URLs, script tags, or JSON-LD
• Regex Extractor — Pattern-based extraction with built-in, custom, or LLM-generated patterns
• Cosine Similarity — Semantic clustering (requires unclecode/crawl4ai:all Docker image)
• SEO Metadata — Meta tags, Open Graph, Twitter Cards, JSON-LD, robots, hreflang

Jobs & Monitoring

• Submit Crawl Job — Async crawl via /crawl/job with webhook support
• Submit LLM Job — Async LLM extraction via /llm/job
• Get Job Status — Poll /job/{task_id} for results
• Health Check — Server health and endpoint stats

Requirements

• n8n: 1.79.1 or higher
• Crawl4AI Docker: 0.8.0
  • Standard operations: unclecode/crawl4ai:latest
  • Cosine Similarity Extractor: unclecode/crawl4ai:all (includes sentence-transformers)

Installation

Via n8n UI (recommended)

1. Go to Settings → Community Nodes
2. Click Install a community node
3. Enter n8n-nodes-crawl4ai-plus
4. Restart n8n

From source (development)

# Install with pnpm (required — npm/yarn not supported)
pnpm install
pnpm build

Then restart your n8n instance. The nodes are declared in package.json → "n8n" → "nodes" and loaded from dist/.

Setup

Credentials

1. Settings → Credentials → New → Crawl4AI API
2. Configure:
   • Docker URL — URL of your Crawl4AI container (default: http://crawl4ai:11235)
   • Authentication — Defaults to No Authentication, which is correct for a standard Docker quickstart deployment. Switch to Token or Basic auth only if your Crawl4AI instance is configured with authentication.
   • LLM Settings — Enable and configure a provider for AI-powered operations:
     • OpenAI, Anthropic, Groq, Ollama, OpenRouter, or Custom (OpenAI-Compatible) for any OpenAI-compatible API (LLM7, OpenCode, Together, Fireworks, etc.)

Custom (OpenAI-Compatible) Provider

The Custom provider option works with any API that implements the OpenAI chat completions format. This is useful for services like LLM7, OpenCode, Together AI, Fireworks AI, or self-hosted models behind a compatible proxy.

To configure:

1. Set LLM Provider to Custom (OpenAI-Compatible)
2. Enter the Base URL — the full API endpoint (e.g. https://api.llm7.ai/v1, https://opencode.ai/zen/v1)
3. Enter the Model Name — the model identifier your API expects (e.g. mistral-7b)
4. Enter the API Key (optional — leave empty if the API doesn't require one)

How it works: Crawl4AI's Docker container uses litellm internally for LLM routing. The Custom option uses litellm's openai/ provider prefix with a custom base_url, which routes requests to your endpoint instead of api.openai.com. This is litellm's designed-in mechanism for supporting arbitrary OpenAI-compatible APIs.

Simple Node

1. Add Crawl4AI Plus to your workflow
2. Select an operation (Get Page Content, Ask Question, Extract Data, or CSS Extractor)
3. Configure the URL and required fields
4. Optional settings are in a single flat Options collection

Advanced Node

1. Add Crawl4AI Plus Advanced to your workflow
2. Select an operation from one of the 3 groups (Crawling, Extraction, Jobs & Monitoring)
3. Configure the URL/required fields
4. Fine-tune via 3 standardized collections: Browser & Session, Crawl Settings, Output & Filtering
5. LLM-based operations use the provider configured in credentials

Configuration Reference

Browser Options

Session & Authentication

Crawler Options

Deep Crawl Options (Crawl Multiple URLs — Discover mode)

Output Options

Output Shape

All operations return a consistent output object:

{
  "domain": "example.com",
  "url": "https://example.com/page",
  "fetchedAt": "2026-02-18T10:00:00.000Z",
  "success": true,
  "statusCode": 200,
  "content": {
    "markdownRaw": "...",
    "markdownFit": "..."
  },
  "extracted": {
    "strategy": "JsonCssExtractionStrategy",
    "json": { ... }
  },
  "links": {
    "internal": [{ "href": "...", "text": "..." }],
    "external": []
  },
  "metrics": {
    "durationMs": 1240
  }
}

Async Job Workflow

For large or long-running crawls, use the async pattern:

1. Submit Crawl Job → returns taskId
2. Get Job Status (poll with taskId) → returns status: pending | processing | completed | failed
3. When completed, result fields are returned directly at top level alongside taskId and status

Webhook callbacks are supported in Submit Crawl Job for push-based notification when the job finishes.

Project Structure

nodes/
  ├── shared/                                 # Shared code used by both nodes
  │   ├── apiClient.ts                        # Crawl4aiClient — all HTTP calls
  │   ├── utils.ts                            # Config builders, LLM helpers, validation
  │   ├── interfaces.ts                       # TypeScript types
  │   ├── formatters.ts                       # formatCrawlResult, formatExtractionResult
  │   └── descriptions/                       # Reusable n8n UI field definitions
  │       ├── index.ts                        # Barrel export
  │       ├── common.fields.ts                # urlField, urlsField, cacheModeField, etc.
  │       ├── browserSession.fields.ts        # getBrowserSessionFields()
  │       ├── crawlSettings.fields.ts         # getCrawlSettingsFields()
  │       └── outputFiltering.fields.ts       # getOutputFilteringFields()
  │
  ├── Crawl4aiPlus/                           # Simple node (4 operations)
  │   ├── Crawl4aiPlus.node.ts
  │   ├── crawl4aiplus.svg
  │   ├── actions/
  │   │   ├── operations.ts
  │   │   ├── router.ts
  │   │   ├── getPageContent.operation.ts
  │   │   ├── askQuestion.operation.ts
  │   │   ├── extractData.operation.ts
  │   │   └── cssExtractor.operation.ts
  │   └── helpers/
  │       ├── utils.ts                        # getSimpleDefaults, executeCrawl, deduplicateResults
  │       └── formatters.ts                   # Simple node formatters
  │
  └── Crawl4aiPlusAdvanced/                   # Advanced node (15 operations, 3 groups)
      ├── Crawl4aiPlusAdvanced.node.ts
      ├── crawl4aiplus.svg
      ├── actions/
      │   ├── operations.ts                   # 15 operations with groupName for UI grouping
      │   ├── router.ts
      │   ├── crawlUrl.operation.ts           # ─┐
      │   ├── crawlMultipleUrls.operation.ts  # │ Crawling group
      │   ├── crawlStream.operation.ts        # │
      │   ├── processRawHtml.operation.ts     # │
      │   ├── discoverLinks.operation.ts      # ─┘
      │   ├── llmExtractor.operation.ts       # ─┐
      │   ├── cssExtractor.operation.ts       # │
      │   ├── jsonExtractor.operation.ts      # │ Extraction group
      │   ├── regexExtractor.operation.ts     # │
      │   ├── cosineExtractor.operation.ts    # │
      │   ├── seoExtractor.operation.ts       # ─┘
      │   ├── submitCrawlJob.operation.ts     # ─┐
      │   ├── submitLlmJob.operation.ts       # │ Jobs & Monitoring group
      │   ├── getJobStatus.operation.ts       # │
      │   └── healthCheck.operation.ts        # ─┘
      └── helpers/
          ├── interfaces.ts                   # Re-exports shared types
          └── formatters.ts                   # Re-exports shared + formatJobSubmission()

credentials/
  └── Crawl4aiApi.credentials.ts             # Docker URL, auth, LLM provider config

Version History

See CHANGELOG.md for detailed version history and breaking changes.

License

MIT