Firecrawl and OpenCode: MCP setup and scraping workflows

This tutorial connects Firecrawl to OpenCode through remote MCP and uses it for web research and automation. The presenter starts in the playground with search, Markdown extraction, URL mapping and crawling, then authorizes the connection inside OpenCode. Alexandria providers extend the sources an AI agent can access.

The examples cover tool discovery, manufacturer lead generation and freelance project ideas drawn from public business reviews. The discovery workflow produces a narrated video using Fish Audio. The lead workflow selects 20 companies across Texas and Louisiana, checks their locations, removes duplicates and flags missing data. The presenter recommends completing a task manually and refining its steps before turning it into a reusable skill.

The self-hosted section explains the limits of the default Firecrawl stack. According to the guide discussed by the presenter, core scraping, crawling, mapping and search use the default stack, while LLM extraction needs a compatible provider or Ollama and separate testing. Fire Engine requires separate configuration; screenshots and page actions depend on it. Browser interaction and specialized formats may require Firecrawl Cloud or external services. Running the core on your own network therefore does not establish feature parity with the cloud service.