Today, we're launching Firecrawl CLI and Skill - the complete web data toolkit for AI agents. Scrape, search, and browse the web, all from a single unified interface.

• Firecrawl CLI is a unified command-line interface for scraping, searching, and browsing the web - with results written directly to the filesystem.
• The Firecrawl Skill teaches AI agents how to install and use the CLI on their own - no manual setup required.
• Together, they give agents clean, structured web data exactly when they need it.

Just run:

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

That's it. This installs the CLI, authenticates through the browser, and adds the Firecrawl skill to all your agents.

Most agents rely on web fetch tools that break on JavaScript-heavy sites, miss key content, or return unusable HTML. When they do get data, they dump entire pages into context, wasting tokens and slowing down reasoning.

The Firecrawl CLI and Skill gives AI agents a complete web data toolkit they can use on their own. With a single install, agents like Claude Code, Antigravity, and OpenCode get a unified interface for scraping pages, searching the web, and browsing interactive sites - plus crawling and mapping for structured site exploration.

Firecrawl CLI is a unified command-line interface designed for agents and developers who need reliable web data.

• scrape - Pull clean markdown from any page, even JavaScript-heavy sites
• search - Search the web and scrape the results in one step
• browser - Launch persistent or temporary cloud browser sessions to interact with any site - click, fill forms, take snapshots, and extract data
• crawl - Follow links recursively and gather content from an entire site
• map - Discover all URLs on a domain for comprehensive coverage

Firecrawl CLI uses a file-based approach for context management and bash methods for efficient search and retrieval.

Firecrawl delivers >80% coverage on benchmark evaluations - outperforming every other provider tested. It handles JavaScript rendering, dynamic content, authentication flows, and complex page structures that break other scrapers.

(Methodology: Checked 1,000 URLs for content recall and whether various tools in the market retrieved at least 10% of the expected content gap, using the Firecrawl scrape-content-dataset-v1 dataset.)

Example usage:

# Scrape a page to markdown
$ firecrawl scrape https://example.com/pricing --format markdown -o pricing.md
 
# Search and scrape top results
$ firecrawl search "AI agent benchmarks 2026" --scrape --limit 5 -o results/
 
# Browse an interactive site with a cloud browser
$ firecrawl browser launch-session
$ firecrawl browser execute "open https://news.ycombinator.com"
$ firecrawl browser execute "snapshot"
$ firecrawl browser execute "click @e5"
$ firecrawl browser execute "scrape"
$ firecrawl browser close
 
# Crawl an entire website
$ firecrawl crawl https://example.com --wait --progress -o example-crawl.json
 
# Map an entire domain
$ firecrawl map https://example.com -o sitemap.json

All results are written to the filesystem, so agents can search, analyze, or process data locally.

The Firecrawl Skill is a skill package that teaches AI agents how to install, authenticate, and use Firecrawl CLI - end to end.

Skills are declarative packages that agents can install via the npx skills protocol. Once installed, agents know how to use the tool without manual integration.

1. How to install Firecrawl CLI
2. How to authenticate with an API key
3. When and how to use scrape, search, and browser commands
4. How to structure output for efficient filesystem usage

The Firecrawl Skill works with any agent that supports the Skills protocol:

• Claude Code - Agentic coding in your terminal
• OpenAI Codex - Build with GPT-4 in the CLI
• Gemini CLI - Google's agent interface
• OpenCode - Open-source agent harness
• And more

One skill, every major coding agent harness.

You don't configure endpoints, handle auth, or write tool definitions. The Skill does it. Add Firecrawl once, and your agent knows how to use it forever.

As the Skill evolves, your agent learns new capabilities automatically. No code changes. No redeployment. It just works. If you want to build your own Claude Code automation skill for a custom tool or workflow, we have a step-by-step tutorial that covers the full structure from scratch.

When you combine Firecrawl CLI with our skill, your agent can reliably access web data on its own.

1. You run $ npx -y firecrawl-cli@latest init --all --browser
2. The skill teaches your agent about Firecrawl CLI capabilities
3. When your agent needs web data, it installs and uses the CLI automatically
4. Results are written to the filesystem, not dumped into context
5. Your agent searches, analyzes, or processes the data locally

Here's a quick look at common use cases where agents with Firecrawl shine:

Install the Firecrawl CLI and skill in all your agents:

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

Or use the CLI directly:

$ npm install -g firecrawl-cli
$ firecrawl auth --api-key fc-YOUR_KEY
$ firecrawl scrape https://example.com

Check out our docs for more info.

Give your agent the best web data available. No context bloat. Just reliable, real-time information when it needs it.

Resources:

• GitHub: github.com/firecrawl/cli
• Documentation: docs.firecrawl.dev/cli
• Get an API key: firecrawl.dev