Skip to main content
Build a tool layer that lets AI agents — Claude Desktop, ChatGPT, LangChain chains, or autonomous agents — discover products, execute purchases, and settle payments in USDC. No human in the loop required.

What You’ll Build

Prerequisites

Pattern 1: Build an MCP Server

Expose Podium operations as Model Context Protocol tools that Claude Desktop (or any MCP client) can call.

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

Pattern 2: A2A Discovery

Expose your commerce capabilities via the Agent-to-Agent protocol discovery endpoint. Create /.well-known/agent.json:

Pattern 3: Agentic Product Feed + Checkout

The agentic endpoints are designed for agent consumption — no session management, no auth cookies, just API key + structured responses.

Browse Products

Create a Checkout Session

Update Session (Add Email, Change Address)

Check Session Status

Pattern 4: x402 USDC Payment

The x402 protocol enables machine-native payments. The agent pays with USDC — no card, no human approval.

Pattern 5: LangChain Tools

Wrap the SDK as typed LangChain tools for reliable agent chains: