MCP-Crawl4AI

Architecture

Canonical architecture for MCP-Crawl4AI

Architecture

MCP-Crawl4AI uses a single-module FastMCP server backed by one lifespan-managed AsyncWebCrawler.

Canonical tool surface: scrape, crawl, close_session, and get_artifact.

System Overview

Rendering diagram...
High-level architecture of MCP-Crawl4AI

Core Components

FastMCP v3 Server

Handles transport, MCP protocol plumbing, and registration for tools/resources/prompts.

Lifespan Manager

Initializes and tears down a shared AsyncWebCrawler once per server lifecycle.

Canonical Tool Surface

ToolPurpose
scrapeCanonical single/batch page scraping envelope
crawlCanonical list/deep traversal envelope
close_sessionExplicit session shutdown
get_artifactArtifact retrieval for captured runs

Resource Surface

Resource URIDescription
config://serverServer configuration, capabilities, and tool inventory
crawl4ai://versionServer, crawl4ai, and fastmcp versions

Data Flow

Rendering diagram...
Canonical scrape/crawl execution flow

Found an issue with this page? Submit a GitHub issue

On this page