Search skills/

Firecrawl CLI Claude skills

Firecrawl's command-line client and 13 agent skills for web search, scraping, crawling, interacting with pages and parsing files through the paid Firecrawl API.

firecrawl/cli 13 skills ResearchBrowser automation & scraping
Our verdict Recommended with caveats

Capable and clearly documented, but it is a front end for a paid hosted service with credit costs and anonymous auth-time telemetry that you may want to switch off.

2/5 Depth expertise shipped Note Security Review before installing 1d Maintenance Updated today 660K Adoption installs ? License no clear license

What Firecrawl CLI does

The Firecrawl CLI lets an agent do live-web work from the shell and returns clean markdown meant to fit a context window. The commands are search, scrape, map (list a site's URLs), crawl, download, interact (click and log in on pages), parse (local files), monitor (change alerts) and agent (multi-minute extraction jobs). Output goes to files under .firecrawl/, which the install guide says is added to .gitignore, and the security rules tell the agent to read results incrementally and treat fetched pages as untrusted.

The repo's skills folder holds 13 skills averaging about 6 KB: one general firecrawl skill that routes web tasks, a skill per command, a monitor skill with goal and JSON-tracking references, a developer-index skill for library and API questions, a research-index skill for papers, and an Alexandria skill for ready-made data providers. Setup is firecrawl setup core, build or workflows, or a single skill by name, and the init command installs into every detected agent, including Claude Code, Codex, Cursor, Windsurf, OpenCode, OpenClaw, OpenHands and Hermes Agent. Everything runs against Firecrawl's hosted API with credits, or a self-hosted instance if you point the CLI at your own URL.

Good for

  • Giving an agent reliable web search and page-to-markdown extraction
  • Bulk crawling or mapping a site for research, SEO audits or knowledge bases
  • Pages that need clicks or login via the interact command

Not for

  • Fully offline or no-account workflows (the hosted API needs a key beyond a limited keyless tier)
  • People who want a free, local-only scraper

What's inside depth 2/5

13 skills averaging 6 KB with a few small rule or reference files, no scripts or tests in the repo; the real functionality is in the CLI and hosted API.

Skills
13
Reference docs
6
Scripts
0
Tests / evals
0
Templates
0
Avg SKILL.md
6.1 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Node.js/npm to install firecrawl-cli
  • Firecrawl account and API key (credits are consumed; the skill mentions a keyless free tier)
  • Optional: self-hosted Firecrawl instance via a custom API URL
  • Telemetry: Per the README, the CLI sends anonymous usage data during authentication (CLI version, OS, Node.js version and detected dev tools such as Cursor or Claude Code); it says no command data, URLs or file contents are collected. Disable with FIRECRAWL_NO_TELEMETRY=1. Update checks are disabled with FIRECRAWL_NO_UPDATE_CHECK=1, and Alexandria feedback calls with FIRECRAWL_NO_ENDPOINT_FEEDBACK=1.

Security review

Review before installing

The Firecrawl web-scraping CLI and its skills. It installs a binary, stores your Firecrawl key in a user-only file, sends URLs and queries to the Firecrawl API, and can write MCP and skill configuration into your coding agents. Login sends small usage metadata unless you opt out, and the skills pre-approve the firecrawl CLI (and its npx form) in allowed-tools.

Worth knowing

  • Installs with curl | shDocumented install is curl -fsSL https://firecrawl.dev/install.sh | bash (vendor host); the script downloads a release binary from GitHub, verifies its sha256 against checksums.txt when one is found, installs to ~/.local/bin and forwards extra args to firecrawl init.scripts/install.sh:5
  • Changes files outside your projectAppends a PATH export to your shell rc file; firecrawl setup/init write MCP and skill configuration into other agents' config files (for example ~/.hermes/config.yaml) and install skills into agent folders; credentials are saved to ~/.config/firecrawl-cli/credentials.json (mode 0600).scripts/install.sh:124
  • Needs an API key for its own serviceFIRECRAWL_API_KEY (or the stored key from browser login) is sent only to https://api.firecrawl.dev, or to the URL set in FIRECRAWL_API_URL for self-hosting; MCP setup writes an environment-variable reference rather than the key.src/utils/config.ts:32
  • Sends usage data by default (can be turned off)During browser login the CLI adds base64 metadata (CLI version, OS, Node version, detected coding agents) to the login URL; disable with FIRECRAWL_NO_TELEMETRY=1. An update-check against the npm registry and the skills repo runs unless FIRECRAWL_NO_UPDATE_CHECK is set.src/utils/auth.ts:566
  • Makes network requestsScrape, crawl, search, parse, browser and feedback commands send URLs, queries, uploaded files or feedback to api.firecrawl.dev; firecrawl terms and the MCP setup use mcp.firecrawl.dev and firecrawl.dev.src/utils/client.ts:55
  • Installs packages or toolsfirecrawl create/init run npm install (or the detected package manager) and git clone --depth 1 of template and skill repositories from GitHub.src/utils/agent-scaffold/scaffold.ts:235
  • Runs remote code at an unpinned versionREADME and skills run npx -y firecrawl-cli and npx skills without a version.README.md:14

What we checked

The repository (165 files) on Oct 29, 2026. Not a full review: we read 2 of 130 code files; it relies on code outside the repository: npm package firecrawl-cli as published, the prebuilt binaries from GitHub releases, and the Firecrawl hosted API / MCP service; Not read in full: src/index.ts and the other command, util and type files, scripts/install.ps1 and build scripts, and the test suite (inspected only by grep for network calls, child processes, file writes and credential use); Skill repositories cloned at install time (firecrawl/skills and chosen templates).

NVIDIA SkillSpector static scan: 168 pattern hits (MCP Rug Pull 54, Tool Misuse 34, Data Exfiltration 28, Privilege Escalation 24, Supply Chain 21, Excessive Agency 4). Most hits are unpinned npx mentions, test fixtures with fake keys, .env and credentials.json handling in tests, and CI workflow lines (deploy key, persist-credentials). 'Collect all env vars' is the agent-scaffold wizard prompting the user for values to write to a new project's .env. The skills' security rule treats fetched web content as untrusted data.

A review, not a certification. Skills change after we check them — read the files before installing.

License

No license detected in our copy of the repo — all rights reserved by default until the authors state otherwise.

Can it run unattended?

Yes. The monitor and crawl commands run as long jobs with credit limits (--max-credits) and file output, though success checking is left to the agent.

Skills in this repository

SkillInstalls
firecrawl104K
firecrawl-search85K
firecrawl-scrape84K
firecrawl-agent81K
firecrawl-crawl81K
firecrawl-map80K
firecrawl-download80K
firecrawl-interact65K

Linked skills have their own review.

Alternatives to Firecrawl CLI

Firecrawl Skills

firecrawl/skills

Firecrawl's official catalog of 37 skills for web search, scraping, crawling and browser interaction, plus API integration and workflow recipes.

With caveats Check first Depth 3/5

agent-browser

vercel-labs/agent-browser

A native browser-automation CLI for agents, plus 11 skills that teach the snapshot-and-ref workflow for navigating, filling forms, scraping and testing.

Recommended Check first Depth 4/5

Apify Agent Skills

apify/agent-skills

Apify's official skills for scraping with its Actor marketplace and for building, converting and integrating Actors with the apify CLI.

With caveats Check first Depth 3/5

Browser Use

browser-use/browser-use

Seven skills around the Browser Use browser agent: drive local Chrome over CDP, use the paid cloud API, run QA scoring, or pay by crypto wallet.

With caveats Broad access Depth 3/5

FAQ

What is the Firecrawl CLI?

It is the command-line client for Firecrawl's web API, plus a set of agent skills. It lets coding agents search the web, scrape pages into markdown, crawl and map sites, interact with pages and run longer extraction jobs.

How do I install Firecrawl skills for Claude Code?

Run npx -y firecrawl-cli@latest init -y --browser to install the CLI, sign in and add skills for the coding agents it detects. Or run npm install -g firecrawl-cli and then firecrawl setup core, optionally with --agent claude-code.

Is the Firecrawl CLI free and safe?

The CLI itself is free to install but uses Firecrawl's hosted API, which spends credits after any free tier. We found no hidden code in the repo; it documents anonymous telemetry at login that you can disable with FIRECRAWL_NO_TELEMETRY=1.

Firecrawl vs agent-browser: which should I use?

Firecrawl is a hosted service optimized for search, scraping and crawling into clean markdown at scale. agent-browser is a local CLI that drives a real Chrome on your machine. Use Firecrawl for bulk extraction and research, and agent-browser for testing and local logged-in sessions.

Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026); install counts from skills.sh via Best Skills (CC BY 4.0). Review based on the README and SKILL.md files as published.