@parke.dev/pi-firecrawl

Firecrawl search, scrape, map and crawl tools for the pi coding agent

Packages

Package details

extensionskill

Install @parke.dev/pi-firecrawl from npm and Pi will load the resources declared by the package manifest.

npm repo home report

$ pi install npm:@parke.dev/pi-firecrawl

Package: @parke.dev/pi-firecrawl
Version: 0.1.0
Published: Aug 11, 2026
Downloads: 104/mo · 8/wk
Author: lukehagar
License: MIT
Types: extension, skill
Size: 23.3 KB
Dependencies: 1 dependency · 3 peers

Pi manifest JSON

{
  "extensions": [
    "./extensions/firecrawl.ts"
  ],
  "skills": [
    "./skills"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

@parke.dev/pi-firecrawl

Firecrawl tools for the pi coding agent. Turn web pages into clean markdown, search the web, map a site's URLs, or crawl many pages in one job — against Firecrawl's hosted API or a self-hosted v1 instance.

Four tools:

Tool	What it does
`firecrawl_scrape`	scrape one URL into markdown (and optional other formats)
`firecrawl_search`	web search; optionally scrape each result page
`firecrawl_map`	discover the URLs on a site
`firecrawl_crawl`	crawl linked pages into markdown (async job, polls to completion)

Install

pi install npm:@parke.dev/pi-firecrawl

By default the tools talk to the hosted API at https://api.firecrawl.dev, which needs an API key and consumes credits. Point baseUrl at a self-hosted instance instead and no key is required on a typical local setup. Run /firecrawl in pi to confirm connectivity.

Configuration

Everything is optional. Precedence is defaults ← ~/.pi/firecrawl.json ← environment.

{
  "baseUrl": "https://api.firecrawl.dev",
  "apiKey": "…",
  "timeoutMs": 120000,
  "crawlTimeoutMs": 120000
}

Field	Env var	Default	Meaning
`baseUrl`	`FIRECRAWL_BASE_URL`	`https://api.firecrawl.dev`	Firecrawl API origin
`apiKey`	`FIRECRAWL_API_KEY`	none	sent as `Authorization: Bearer`; required by the hosted API
`timeoutMs`	`FIRECRAWL_TIMEOUT_MS`	120000	request timeout for scrape / search / map
`crawlTimeoutMs`	`FIRECRAWL_CRAWL_TIMEOUT_MS`	120000	how long to poll a crawl job before giving up

Deployment shapes

Hosted API (the default) needs a key and burns credits:

{ "apiKey": "fc-…" }

Or export FIRECRAWL_API_KEY=fc-….

Self-hosted only needs the origin:

{ "baseUrl": "http://localhost:3002" }

Diagnostics

/firecrawl    show config and test connectivity with a scrape of example.com

License

MIT