An AI-powered browser automation skill for Openclaw Skills that records interactions as trajectories and replays them as low-token scripts.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install zerotoken
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 zerotoken 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 ZeroToken OpenClaw?
ZeroToken is a specialized Model Context Protocol (MCP) integration designed to bring high-efficiency browser automation to the Openclaw Skills ecosystem. It empowers AI agents to record complex web interactions as trajectories and subsequently replay them using a dedicated script engine. This architecture significantly reduces token consumption by bypassing the need for the Large Language Model to process every intermediate visual step during repetitive tasks.
By leveraging ZeroToken within the framework of Openclaw Skills, developers can build robust, stealth-enabled automation workflows. The skill excels at handling everything from simple form filling to complex multi-page data extraction. It is specifically optimized for environments where state management is critical, utilizing a persistent HTTP transport layer to ensure browser sessions remain active across multiple agent calls.
ZeroToken OpenClaw Use Cases
- Automating recurring daily logins and report generation on enterprise web platforms.
- Recording complex multi-step browser trajectories for future low-token replay and cost optimization.
- Executing scheduled browser tasks triggered by specific job IDs within the Openclaw Skills environment.
- Handling data extraction from websites with sophisticated anti-bot protection using integrated stealth mode.
- Transforming manual browser exploration sessions into reusable, parameterized automation scripts.
How ZeroToken OpenClaw Works
- Initialize the browser session using the browser_init tool, with the option to enable stealth mode for anti-detection.
- Start a recording session via trajectory_start to capture all subsequent navigation, clicks, and inputs.
- Interact with the target website using standard browser tools like browser_click and browser_input.
- Finalize the recording with trajectory_complete to store the interaction sequence in the local SQLite database.
- Convert the captured trajectory into a reusable script using trajectory_to_script, which optimizes the sequence for low-token execution.
- Execute the automation autonomously using run_script or bind it to a job_id for scheduled workflows.
ZeroToken OpenClaw Setup
To successfully integrate this skill with Openclaw Skills, you must configure the ZeroToken MCP server to run in HTTP mode to maintain persistent browser state.
- Install the core package and browser dependencies:
pip install zerotoken
playwright install chromium
- Launch the ZeroToken MCP server in the background using the streamable HTTP transport:
zerotoken-mcp-http --port 8000
- Update your openclaw.json configuration to connect to the running server:
{
"mcpServers": {
"zerotoken": {
"url": "http://localhost:8000/mcp"
}
}
}
- Deploy the skill folder into your Openclaw local or workspace skills directory.
ZeroToken OpenClaw Data Schema & Taxonomy
ZeroToken organizes its automation data within a structured SQLite database managed by the MCP server, ensuring that Openclaw Skills do not need to manage local file paths directly.
| Entity | Description |
|---|---|
| Trajectories | Raw recordings of browser interactions including selectors, actions, and metadata. |
| Scripts | Refined sequences of MCP tool calls optimized for autonomous replay and low token usage. |
| Sessions | Historical logs of every recording or replay attempt, used for auditing and debugging. |
| Bindings | A mapping table that links specific Openclaw job IDs to their corresponding automation scripts. |
ZeroToken OpenClaw Advanced Features
- Stealth Mode: Sophisticated fingerprint spoofing including navigator disguise, WebGL masking, and Sec-CH-UA header injection to bypass anti-bot systems.
- Parameterized Execution: Support for {{varname}} placeholders within scripts, allowing Openclaw Skills to inject dynamic data at runtime.
- Job ID Integration: Native support for run_script_by_job_id, allowing scheduled tasks to automatically find and trigger the correct automation script.
- Fuzzy Point Identification: Semantic markers within scripts that identify steps requiring AI decision-making or human intervention.
- Adaptive Element Locating: Self-correcting selector logic that maintains script stability even when website UI elements undergo minor changes.
SKILL.md
Loading
Related Openclaw Skills
Qualify
A strategic decision-support tool that defines and applies qualification logic to filter out noise and focus resources on high-value opportunities.
Organizer
A professional framework for turning cluttered information into structured, navigable, and actionable systems.
AetherCore
A security-focused high-performance JSON optimization and universal smart indexing engine designed for Openclaw Skills.
Sensitive Content Detection API
A robust FastAPI-based service for identifying sensitive words, typos, and linguistic inconsistencies in text.
OpenClaw Security Policy Check
An automated security auditing tool designed to identify and remediate configuration risks in OpenClaw gateway environments.
Shopify Store Manager CLI
A Python-based CLI tool to manage Shopify store resources including products, metaobjects, and content through the official GraphQL API.
Featured*
