Deploy and Host Web Tools on Railway

Web Tools is an open-source web toolkit that gives AI agents eight tools to search, fetch, screenshot, crawl, and archive the web. Available as an MCP server, REST API, and CLI. It consumes zero LLM tokens for web access, so your models spend their budget on reasoning, not searching. The web has always been free for humans, so why should AI agents have to pay per query?

About Hosting Web Tools

This template deploys a complete self-hosted web toolkit as four services on Railway: Redis, SearXNG (privacy-respecting metasearch engine), Crawl4AI (headless browser for content extraction, screenshots, PDFs, and JS execution), and the Web Tools Server that ties them together. An API key is auto-generated at deploy time to secure your endpoint. Once deployed, any MCP-compatible client (Claude Code, Claude Desktop, Cursor, Windsurf, etc.) can connect over HTTP and use all eight tools. A REST API (POST /api/v0/{tool_name}) is also available for non-MCP integrations. No per-query fees, no third-party API keys, no usage limits. You own the infrastructure and the data never leaves your stack.

Common Use Cases

Replace paid search APIs: Drop-in replacement for Firecrawl, Linkup, Tavily, Exa, or Bright Data. Get web search, page fetching, and content extraction without per-query costs
Supercharge AI coding agents: Connect Claude Code or Cursor to self-hosted web search and page fetching. Replace their built-in WebSearch and WebFetch tools so every search is private and free
Web research and monitoring: Search the web, fetch pages as clean markdown, take screenshots, generate PDFs, execute JavaScript on pages, and query the Wayback Machine for historical snapshots
Build custom integrations: Use the REST API to integrate web tools into any application or workflow

Dependencies for Web Tools Hosting

Redis (7-alpine): In-memory cache used by SearXNG for rate limiting and result caching
SearXNG: Privacy-respecting metasearch engine that aggregates results from Google, Brave, DuckDuckGo, and more. Builds from services/searxng/Dockerfile with optional PROXY_URL support for outgoing requests
Crawl4AI: Headless browser service for page fetching, content extraction, screenshots, PDFs, and JavaScript execution
Web Tools Server (Node.js 22): The HTTP server exposing MCP and REST API endpoints. Builds from packages/api/Dockerfile

Deployment Dependencies

Implementation Details

The Web Tools Server exposes two interfaces:

MCP — Streamable HTTP endpoint at /mcp for MCP clients:

{
  "mcpServers": {
    "web_tools": {
      "type": "http",
      "url": "https://your-server.up.railway.app/mcp",
      "headers": {
        "Authorization": "Bearer your-api-key"
      }
    }
  }
}

REST API — Standard HTTP endpoints at /api/v0/{tool_name}:

curl -X POST https://your-server.up.railway.app/api/v0/web_search \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"query": "railway deployment"}'

The eight tools available are: web_search, web_fetch, web_screenshot, web_pdf, web_execute_js, web_crawl, web_snapshots, and web_archive.

Railway Service Configuration

Service	Source	Root Directory	Notes
Web Tools Server	GitHub repo	(repo root)	Uses root `Dockerfile`, exposes MCP + REST API
SearXNG	GitHub repo	`services/searxng`	Optional `PROXY_URL` env var
Crawl4AI	Docker image (`unclecode/crawl4ai:latest`)	—
Redis	Docker image (`redis:7-alpine`)	—

Why Deploy Web Tools on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Web Tools on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.