
AgentCard
@agent-cards
AgentCard について
MCP server for Agent Cards — prepaid virtual Visa cards for AI agents
基本情報
設定
以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。
{
"mcpServers": {
"agent-cards": {
"url": "https://mcp.agentcard.sh/mcp",
"headers": {
"Authorization": "Bearer <your-jwt>"
}
}
}
}ツール
16List all virtual cards with balances, expiry, and status
Create a new virtual debit card with a fixed USD budget
Get decrypted PAN, CVV, expiry (may require human approval)
Fast balance check without exposing credentials
Permanently close a virtual card
List transactions for a card with optional status filter
Approve or deny a pending human-in-the-loop request
Submit identity info for card issuance verification
Set up a payment method for funding cards
Remove a saved payment method
Auto-detect and pay a checkout page using an AgentCard
Detect checkout forms on the current browser tab
Fill card credentials into a payment form in the browser
Start a support conversation
Send a message in a support thread
Read support conversation history
概要
What is AgentCard?
AgentCard lets AI agents create and spend virtual debit cards, each with a fixed budget, real card credentials, and full MCP integration. Your agent can create a card, pay for things, check its balance, and auto-fill checkout forms without ever needing your personal card.
How to use AgentCard?
The fastest setup is via the CLI: run npx agent-cards signup (one-time) then npx agent-cards setup-mcp to auto-configure Claude Code. Alternatively, you can manually add a Streamable HTTP or stdio MCP server entry to your client config, providing your JWT token via header or environment variable.
Key features of AgentCard
- Create virtual debit cards with a fixed USD budget
- Retrieve decrypted PAN, CVV, and expiry (with human approval)
- Auto-detect and pay checkout pages in the browser
- Fill card credentials into payment forms automatically
- List cards, check balances, and list transactions
- Human-in-the-loop approval for sensitive operations
Use cases of AgentCard
- An AI assistant pays for an online service using a dedicated virtual card
- An agent fills checkout forms with its own card credentials, keeping your personal card safe
- A developer gives their AI agent a prepaid budget to purchase tools or subscriptions autonomously
- An agent closes or replaces a compromised virtual card without human intervention
- A support bot starts a human chat when a request requires manual approval
FAQ from AgentCard
How do I get a JWT token to authenticate my MCP client?
Run npx agent-cards login; the JWT is stored in ~/.agent-cards/config.json. Use it in the Authorization header for HTTP mode or as the AGENT_CARDS_JWT environment variable for stdio mode.
What transport protocols does AgentCard support?
AgentCard supports both Streamable HTTP (recommended) and stdio transports. The server auto‑discovers its MCP endpoint via the /.well-known/mcp/server-card.json path.
What environment variables are required?
AGENT_CARDS_API_URL is required for all modes. For stdio mode you must also set AGENT_CARDS_JWT. The PORT variable is optional (defaults to 3002) for HTTP mode.
Are there human‑in‑the‑loop approval steps?
Yes. Getting decrypted card details (get_card_details) may require human approval. Use the approve_request tool to approve or deny pending requests.
How can I contact support through the server?
Use the start_support_chat tool to begin a conversation, then send_support_message and read_support_chat to interact with the support thread.
「AI とエージェント」の他のコンテンツ
1Panel
1Panel-dev🔥 1Panel is a modern, open-source VPS control panel — and the only one with native AI agent support. Run Ollama models, deploy OpenClaw agents, and manage your entire server stack from one clean web interface.
Shell and Coding agent for Claude and other mcp clients
rusiaamanShell and coding agent on mcp clients
MCP Manager for Claude Desktop
zueaisimple web ui to manage mcp (model context protocol) servers in the claude app
MCP Client for Ollama (ollmcp)
joniglHarness the power of local LLMs with this TUI MCP Client for Ollama. Featuring all core MCP primitives (tools, prompts, resources), agent mode, multi-server, model switching, streaming responses, human-in-the-loop, thinking mode, model params config, system prompts, and saved pre
🛡️ A.I.G(AI-Infra-Guard)
TencentA full-stack AI Red Teaming platform securing AI ecosystems via OpenClaw Security Scan, Agent Scan, Skills Scan, MCP scan, AI Infra scan and LLM jailbreak evaluation.
コメント