Skip to main content
Podium is designed to be consumed by agents. Every API endpoint works with bearer-token auth, returns structured JSON, and follows consistent patterns — making it straightforward to expose Podium as tools for any AI system. This section covers how to integrate Podium with the most popular agent clients, coding assistants, and frameworks.

​
MCP Clients

Model Context Protocol (MCP) clients connect to a Podium MCP server you build once and reuse everywhere. See the Build a Claude MCP Server recipe for the full server implementation.

Cursor

MCP server config and .cursor/rules for building on Podium inside Cursor.

Claude Desktop

The canonical MCP setup — configure Claude Desktop to search, buy, and manage with Podium tools.

Windsurf

MCP config and workspace rules for Podium development in Windsurf.

​
CLI Agents

Terminal-based coding agents that can scaffold Podium integrations, build features, and run SDK code.

Claude Code

CLAUDE.md context file and MCP config for Anthropic’s terminal agent.

Codex

OpenAI’s Codex CLI — instructions file and Podium project scaffolding.

​
Agent Frameworks

Native tool definitions for building autonomous agents that use Podium as their commerce backend.

OpenClaw

Plugin and skill integration for the open-source personal AI assistant.

LangChain

DynamicStructuredTool wrappers and LangGraph workflow nodes.

Vercel AI SDK

Tool definitions with Zod schemas for Next.js streaming AI apps.

​
Common Pattern

Regardless of which client or framework you use, the integration pattern is the same:
  1. Define tools that wrap Podium SDK methods
  2. Register them in your agent client or framework
  3. Provide your API key via environment variable
  4. The agent calls tools through natural conversation or autonomous workflows