An interactive command-line utility to control browser sessions, perform page actions with natural language, and scrape dynamic, JavaScript-heavy sites.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install firecrawl-interact
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install firecrawl-interact using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is firecrawl-interact?
The firecrawl-interact tool is a powerful asset in the library of Openclaw Skills, enabling developers to control and interact with a live browser session on any scraped page. Designed as a key fall-back option when traditional static scraping methods fail, this skill enables programmatic navigation, form submission, and content extraction via natural language prompts or custom script configurations.
Rather than manually configuring complex Selenium or Playwright environments, integration with other Openclaw Skills allows developer agents to click elements, fill input fields, handle multi-step checkouts, and navigate infinite scrolling feeds dynamically. It represents a highly flexible way to automate interaction workflows natively.
firecrawl-interact Use Cases
- Automating form-filling tasks, multi-step application registration wizards, and structured login procedures.
- Bypassing static parsing limits by interacting with dynamic menus, clicking dropdown elements, or activating pagination buttons.
- Executing advanced, stateful interactions like handling infinite scroll feeds or toggling JS-rendered tabs.
- Performing authenticated data extraction without triggering bot detection systems repeatedly.
How firecrawl-interact Works
- Scrape First: Run an initial scrape sequence on your target webpage to capture the DOM state and establish a unique session scrape ID.
- Pass Commands: Issue direct interaction commands using plain, natural language guidelines via the
--promptparameter or inline scripts via--code. - Execute Live Actions: The background browser engine receives the commands, translates them into DOM events, and runs them within the active page instance.
- Manage Storage States: Maintain continuous user cookies, session data, and localStorage targets using persistent profile workspace states.
- Clean Resources: Send the stop signal directly to terminate background browser executions and securely free system resources.
firecrawl-interact Setup
To leverage this skill within your local agent toolsets, install the core CLI globally:
npm install -g firecrawl
Begin by starting a target scraping capture which returns a trace-id:
firecrawl scrape "https://example.com/login"
Submit your natural language instructions directly to interact with the target elements:
firecrawl interact --prompt "Type my-username in the username box and click submit"
Release active terminal and browser sessions once your operation completes:
firecrawl interact stop
firecrawl-interact Data Schema & Taxonomy
The utility configures interactive commands and variables using defined parameters. Below is a specification of the schema properties used:
| Command Option | Type | Description | Mandatory |
|---|---|---|---|
--prompt |
String | A descriptive, natural language prompt directing the agent's actions on the page. | Yes (or --code) |
--code |
String | High-precision script blocks to run inside the active browser instance. | Yes (or --prompt) |
--language |
String | The runtime environment for executing custom commands (bash, python, node). | No |
--timeout |
Integer | Session execution runtime limit in seconds (default: 30, max: 300). | No |
--scrape-id |
String | Target trace reference linking this command to the initial scraping phase. | No |
--profile |
String | Tag to persist stateful parameters like authentication tokens and cookies. | No |
-o, --output |
String | The directory path to write the output data results. | No |
firecrawl-interact Advanced Features
- Persistent State Profiles: Track and resume sessions by passing target names with
--profileto manage post-authentication flows smoothly. - Direct Code Injection: Execute precise low-level control code blocks written in Python, Node, or Bash directly within the active web interaction context.
- Read-Only Session Contexts: Reconnect to pre-configured target sessions using
--no-save-changesto prevent temporary testing actions from altering saved cookies. - Multi-Agent Session Management: Map complex processes sequentially across multiple browser tabs by targeting specific scrape IDs.
SKILL.md
Loading
Related Openclaw Skills
firecrawl-crawl
Bulk extract structured content and text from entire websites or targeted sub-sections using customizable depth limits and path filtering.
Firecrawl Scrape
A high-performance web scraping skill that converts static and JavaScript-rendered websites into clean, LLM-optimized markdown.
The Outsiders Skill
An executable capital allocation toolkit modeled after William N. Thorndike's book to help users master strategic financial management, stock buybacks, and rational business decisions.
The Moment of Lift
An executable AI agent skill based on Melinda French Gates' 'The Moment of Lift' that guides strategies on women's empowerment, education, family planning, and grassroots development.
Firecrawl Download
A convenient skill that combines site mapping and scraping to easily download entire websites and documentation locally in markdown, images, or raw text formats.
The Great Displacement Climate Migration Toolkit
An executable AI agent toolkit inspired by Jake Bittle's book to help users understand climate-driven displacement, assess geographic risks, and build resilient adaptation strategies.
Featured*
