article --summary
The summary
Sandboxed coding agents are cut off from documentation sites and recent APIs, so they tend to work from stale memory. This guide shows how to add the Firecrawl kit to a Docker Sandboxes setup so the agent can search, scrape, and crawl the web.
The kit is a mixin that installs the Firecrawl Python SDK, requests a proxy-managed API key, and opens a single network rule for api.firecrawl.dev. You will store the key once on the host, start a sandbox with the kit attached, and run a task that reads a changelog the default policy would block.
The post also walks through checks for the SDK version, the placeholder key, and a live scrape, plus known limits such as blocked media downloads, organization policy overrides, and credit exhaustion. The kit is experimental, so its format and commands may change.
Editorial note: This is a summary of an article first published by Docker Feed. The full piece is on their site.