@firecrawl/opencode

@firecrawl/opencode

OpenCode plugin for Firecrawl. It gives your agent primary-source answers from the Firecrawl developer index, structured data from Firecrawl Alexandria providers, and web scraping, crawling, and search through the Firecrawl CLI.

Installation

The plugin supports OpenCode 2 and OpenCode 1 (1.18.29 or later) from the same package.

OpenCode 2

opencode plugin add @firecrawl/opencode

Or add it to opencode.jsonc yourself:

{
  "$schema": "https://opencode.ai/config.json",
  "plugins": ["@firecrawl/opencode"]
}

OpenCode 1

Add it to opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["@firecrawl/opencode"]
}

OpenCode installs the package on the next start. The developer search tool works right away, with no API key and no CLI. Without a key, requests from some networks (VPNs, proxies, cloud hosts) are refused; connect Firecrawl to search from there.

For the web skills (scrape, crawl, map, search, agent), install the Firecrawl CLI:

npx -y firecrawl-cli@latest init -y --browser

Authentication

On OpenCode 2, run /connect in the TUI (or opencode auth login), pick Firecrawl, and choose Sign in with Firecrawl. OpenCode opens Firecrawl in your browser; approve access and you are connected, with no API key to copy. OpenCode keeps the sign-in fresh on its own. Pasting an API key from firecrawl.dev works too, as does FIRECRAWL_API_KEY in the environment. Connecting or disconnecting takes effect without a restart.

The browser sign-in is used by this plugin's tools and the Firecrawl CLI. OpenCode's built-in Firecrawl web search reads only an API key.

On OpenCode 1, export an API key before starting OpenCode:

export FIRECRAWL_API_KEY=fc-your-api-key

Once connected, the plugin:

• registers the firecrawl_alexandria tool
• passes the credential to shell commands, so the CLI is authenticated
• raises the rate limit on developer search

firecrawl login --browser authenticates the CLI alone.

Tools

firecrawl_developer_search

Searches an index of GitHub issues, merged pull requests, READMEs, and library documentation, and returns the matched passages as markdown rather than a list of links. Use it for how a library behaves, what an error means, whether a bug was fixed, or what an API contract guarantees.

Why does my Playwright script hang on page.goto with a service worker registered?

It takes a query plus optional types (doc, issue, pull_request, readme), repos (owner/name), sources, k, and passages. No API key needed.

firecrawl_alexandria

Finds and runs Alexandria data providers: official APIs, licensed publishers, and Firecrawl indexes that return structured records instead of a web page.

What has the unemployment rate in California been over the last two years?

The agent passes query to discover matching capabilities with their input contract and price (free), then provider, capability, and options to run one. A run spends Firecrawl credits at the listed price. The tool only appears once a key is connected.

Runs go through the firecrawl_alexandria permission, matched on provider/capability, and are allowed unless you configure it. Discovery never asks.

On OpenCode 2 (2.0.29 or later), to confirm each run before it spends:

{
  "permissions": [{ "action": "firecrawl_alexandria", "resource": "*", "effect": "ask" }]
}

Answering "always" remembers that capability. Narrow resource to a provider ("bls-gov/*") to ask for some runs only, and use "effect": "deny" to block them. Earlier OpenCode 2 releases cannot ask from a plugin tool.

On OpenCode 1:

{
  "permission": { "firecrawl_alexandria": "ask" }
}

Skills

The plugin bundles the Firecrawl CLI skills (search, scrape, crawl, map, interact, download, parse, agent, monitor, the developer and research indexes, and Alexandria). They teach the agent when each one fits and how to keep output in a .firecrawl/ directory instead of the context window.

Links

• Firecrawl CLI documentation
• Firecrawl CLI on GitHub
• OpenCode plugin docs

License

ISC