AgentCard is a financial management skill that allows AI agents to create and control prepaid virtual Visa cards for online transactions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-card
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 agent-card 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 AgentCard?
AgentCard is a robust technical solution designed to bridge the gap between AI autonomy and financial transactions. As a key component of Openclaw Skills, it provides a secure interface for AI agents to manage prepaid virtual Visa cards through the agentcard.sh service. This skill enables agents to generate cards with specific spending limits, monitor live balances, and retrieve encrypted credentials for authorized purchases.
By leveraging the Model Context Protocol (MCP), AgentCard ensures that financial operations are handled within a controlled environment. It focuses on security and developer experience, offering a comprehensive suite of tools for identity verification (KYC), transaction history tracking, and automated checkout interactions. This makes it an essential tool for users looking to provide their agents with managed spending capabilities.
AgentCard Use Cases
- Creating restricted-use virtual cards for AI-driven software subscriptions.
- Automating one-time payments for developer tools and cloud services.
- Monitoring and auditing agent spending through real-time transaction logs.
- Facilitating automated checkouts using the integrated Chrome extension support.
- Setting up sandbox environments for testing financial workflows in Openclaw Skills.
How AgentCard Works
- The agent initiates an orientation phase by listing existing cards and identifying current financial resources.
- For first-time users, the skill manages a multi-step onboarding process including KYC submission and Stripe payment method linking.
- Upon user request, the agent creates a virtual card with a defined balance in cents, adhering to safety limits.
- The agent can then monitor transactions or retrieve card details only when explicitly required for a payment task.
- If the AgentCard Pay extension is present, the agent can detect checkout pages and fill form data automatically.
- Finally, the agent can permanently close cards to ensure no further charges are possible, maintaining a clean security posture.
AgentCard Setup
To integrate this functionality into your environment, you must have the agent-cards package installed. Use the following commands to configure the skill for Openclaw Skills:
# Install the agent-cards CLI via npx
npx agent-cards setup-mcp
# Or install globally if preferred
npm install -g agent-cards
After setup, you must restart your agent session to load the new MCP tools. Ensure you complete the KYC identity verification through the submit_user_info tool before attempting to create your first card.
AgentCard Data Schema & Taxonomy
The AgentCard skill organizes data around card entities and transaction histories. The following table describes the primary data points handled by these Openclaw Skills:
| Object | Properties |
|---|---|
| Card | ID, Last Four, Expiry Date, Balance (in cents), Status (ACTIVE/CLOSED) |
| Transaction | Amount, Merchant Name, Status (PENDING/SETTLED), Timestamp |
| Identity | First Name, Last Name, Date of Birth, Phone Number, Terms Acceptance |
| Support | Conversation ID, Message History, Reply Status |
AgentCard Advanced Features
- Sandbox Environment: Support for
sandbox: trueflag to test card creation and workflows without real funds. - Browser Automation: Deep integration with the AgentCard Pay Chrome extension for high-confidence checkout detection.
- Approval Workflows: Built-in handling for 202 Acceptance codes, allowing for human-in-the-loop email approvals for high-value actions.
- Support Ticket Management: Native tools to start, read, and reply to support conversations directly within the agent interface.
- CLI Fallback: A secondary command-line interface for manual management when MCP tools are unavailable.
SKILL.md
Loading
Related Openclaw Skills
Workopia
Workopia is an AI-powered job search, resume generation, and career advisory suite for Openclaw Skills.
GUI Control
Control and automate a Linux GUI desktop environment using command-line tools like xdotool and scrot.
Intake META Ads Analyst
A high-performance analyst skill designed for large-scale META ads management and deep data forensics for the Intake Breathing brand.
NBA Games
A specialized skill for retrieving real-time NBA schedules and historical game results for any professional team.
Seedream 5 via BytePlus Ark
A professional image generation skill for Openclaw Skills that utilizes the Seedream 5 model through the BytePlus Ark API.
会下指令 TaskFit
A local-first prompt optimizer that transforms vague user requests into high-quality, structured execution briefs for AI agents.
Featured*
