Directory MCP server
{
"mcpServers": {
"at-directory": {
"type": "http",
"url": "https://mcp.agenticterminal.ai/mcp"
}
}
}{
"mcpServers": {
"at-directory": {
"type": "http",
"url": "https://mcp.agenticterminal.ai/mcp"
}
}
}{
"mcpServers": {
"at-directory": {
"type": "http",
"url": "https://mcp.agenticterminal.ai/mcp"
}
}
}{
"servers": {
"at-directory": {
"type": "http",
"url": "https://mcp.agenticterminal.ai/mcp"
}
}
}{
"mcpServers": {
"at-directory": {
"type": "http",
"url": "https://mcp.agenticterminal.ai/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.0.0
What it does
AT Directory is an MCP server that indexes merchants accepting Bitcoin Lightning, BOLT12, L402, or USDT payments, with trust attestations from Observer Protocol. It is aimed at AI agents that need to discover and initiate payments with verified merchants. The server sits outside the payment path — agents use it to find merchants and then pay them directly.
- Search for merchants by category or payment rail to find services an agent can pay for autonomously.
- Retrieve a specific merchant's payment endpoint details and verify it is reachable before initiating a transaction.
- Browse available payment rails and merchant categories to decide which integration tier fits a given agent workflow.
Tools this server exposes
| Tool | Does |
|---|---|
| search_merchants | Search the merchant index by keyword, category, or payment rail. |
| get_merchant | Retrieve the full record for a specific merchant by identifier. |
| verify_payment_endpoint | Check that a merchant's payment endpoint is reachable and valid. |
| list_categories | Return the full category taxonomy used to classify merchants. |
| list_rails | Return all supported payment rails and USDT chains in the directory. |
| whoami | Return information about the current caller's identity and access level. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.