agentictrade-mcp 0.1.0

MCP server for AgenticTrade — AI agent API marketplace discovery, service calls, and billing

AgenticTrade MCP Server

An MCP (Model Context Protocol) server that connects AI agents to the AgenticTrade marketplace — discover, call, and pay for API services.

What is AgenticTrade?

AgenticTrade is an API marketplace designed for AI agents. Instead of hardcoding API keys and endpoints, your agent discovers services at runtime, calls them through a payment proxy, and pays per call in USDC. This MCP server exposes the marketplace as tools that any MCP-compatible agent can use.

Tools

Tool	Description	Auth Required
discover_services	Search/browse available API services	No
get_service_details	Get full details of a specific service	No
call_service	Call a service through the payment proxy	Yes
get_balance	Check your agent's USDC balance	Yes
list_categories	List all service categories	No

Installation

pip

pip install agentictrade-mcp

uv (recommended)

uv pip install agentictrade-mcp

From source

git clone https://github.com/judyailab/agentictrade-mcp.git
cd agentictrade-mcp
pip install -e .

Configuration

Environment Variables

Variable	Required	Description
AGENTICTRADE_BASE_URL	No	API base URL (default: https://agentictrade.io)
AGENTICTRADE_API_KEY	For paid calls	Your API key in key_id:secret format
AGENTICTRADE_BUYER_ID	For balance checks	Your buyer/agent ID

Get an API Key

1. Visit agentictrade.io/api-docs
2. Create a buyer key via POST /api/v1/keys
3. Fund your account with USDC, crypto, or PayPal

Usage

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "agentictrade": {
      "command": "agentictrade-mcp",
      "env": {
        "AGENTICTRADE_API_KEY": "your_key_id:your_secret",
        "AGENTICTRADE_BUYER_ID": "your_buyer_id"
      }
    }
  }
}

Cursor

Add to your MCP configuration in Cursor settings:

{
  "agentictrade": {
    "command": "agentictrade-mcp",
    "env": {
      "AGENTICTRADE_API_KEY": "your_key_id:your_secret"
    }
  }
}

Smithery

This server is available on the Smithery registry. Install via:

npx @smithery/cli install agentictrade-mcp

Augment Code

Register as an MCP tool source in Augment Code settings. The server runs on stdio transport and is compatible with any MCP client.

Programmatic (Python)

from agentictrade_mcp.server import mcp

# Run with stdio transport (default)
mcp.run(transport="stdio")

# Or with SSE for remote connections
mcp.run(transport="sse", port=8080)

Tool Details

discover_services

Search the marketplace for APIs your agent can use.

Parameters:
  query        (string)  — Search keywords (e.g., "crypto scanner")
  category     (string)  — Filter by category (e.g., "crypto", "data")
  max_results  (int)     — Max results, 1-100 (default: 20)

Returns: List of services with id, name, description, pricing, quality scores

get_service_details

Get complete information about a specific service.

Parameters:
  service_id  (string)  — Service UUID from discover_services

Returns: Full service info including endpoints, pricing, payment method, tags

call_service

Make a paid API call through the AgenticTrade proxy.

Parameters:
  service_id  (string)  — Service UUID to call
  api_key     (string)  — API key (or set AGENTICTRADE_API_KEY env var)
  payload     (string)  — JSON string of the request body
  path        (string)  — Optional sub-path (e.g., "/api/scan")
  method      (string)  — HTTP method (default: "POST")

Returns: Provider response with billing info (usage_id, amount, latency)

get_balance

Check available funds before making calls.

Parameters:
  api_key   (string)  — API key (or set AGENTICTRADE_API_KEY env var)
  buyer_id  (string)  — Buyer ID (or set AGENTICTRADE_BUYER_ID env var)

Returns: Current balance, total deposited, total spent (all in USDC)

list_categories

Browse what types of APIs are available.

Parameters: none

Returns: List of categories with service counts

Example Agent Flow

1. list_categories()
   → See available categories: crypto, data, ai, ...

2. discover_services(query="crypto price", category="crypto")
   → Find services: CoinSifter Scanner (id: svc-abc, $0.01/call)

3. get_service_details(service_id="svc-abc")
   → Confirm pricing, check free tier, read description

4. get_balance()
   → Balance: $42.50 USDC

5. call_service(service_id="svc-abc", payload='{"symbol": "BTC"}')
   → Response: { "price": 67000, ... }
   → Billing: $0.01 deducted

Development

# Clone and install dev dependencies
git clone https://github.com/judyailab/agentictrade-mcp.git
cd agentictrade-mcp
pip install -e ".[dev]"

# Run tests
pytest

# Run server locally
AGENTICTRADE_BASE_URL=https://agentictrade.io agentictrade-mcp

License

MIT License. See LICENSE.

Links

• Marketplace: agentictrade.io
• API Docs: agentictrade.io/api-docs
• MCP Protocol: modelcontextprotocol.io
• Smithery Registry: smithery.ai
• Built by: JudyAI Lab