Model Context Protocol

Steam MCP Server for AI coding agents

Connect Claude Code, Cursor, Windsurf, Zed or Continue.dev to the Steamwebapi knowledge base. The Steam MCP server gives agents endpoint search, request and response schemas, Steam inventory, item, profile and market concepts, pricing guidance and optional read-only API test calls for integrations using your own Steamwebapi key.

https://mcp.steamwebapi.com/mcp Get an API key

Your API key is shown in the dashboard header, top-right, in the "Your API Key" card.

Install in your editor

Pick the snippet for your client. Project-scoped configs (like .mcp.json in the repo root) let a whole team share the same MCP setup via git.

Claude Code

CLI (user-scope, all projects) 
claude mcp add steamwebapi --transport http --url https://mcp.steamwebapi.com/mcp
Project-scoped — .mcp.json in repo root 
{
  "mcpServers": {
    "steamwebapi": {
      "type": "http",
      "url": "https://mcp.steamwebapi.com/mcp"
    }
  }
}

Cursor

~/.cursor/mcp.json (or Settings → MCP) 
{
  "mcpServers": {
    "steamwebapi": {
      "url": "https://mcp.steamwebapi.com/mcp"
    }
  }
}

Windsurf

Settings → Cascade → MCP Servers 
{
  "steamwebapi": {
    "serverUrl": "https://mcp.steamwebapi.com/mcp"
  }
}

Zed

settings.json 
{
  "context_servers": {
    "steamwebapi": {
      "command": { "path": "https://mcp.steamwebapi.com/mcp" }
    }
  }
}

Continue.dev

~/.continue/config.json 
{
  "experimental": {
    "modelContextProtocolServers": [
      { "transport": { "type": "http", "url": "https://mcp.steamwebapi.com/mcp" } }
    ]
  }
}

What can the Steam MCP server do?

Let an AI coding agent discover the right Steam API endpoint, inspect its schema and produce integration code for inventory, profile, item-price and marketplace workflows.

Tool What it returns
search_endpoints Fuzzy search across all endpoints and Steam domain concepts.
get_endpoint Full OpenAPI spec for a single endpoint: parameters, responses, examples.
list_endpoints Enumerate every endpoint grouped by tag.
list_games Supported Steam games with appId and query slug.
list_markets Third-party marketplaces covered by the price-API.
list_products Product packages and the endpoints each one bundles.
get_concept Domain glossary: assetid, doppler phase, pricelatest, marketable, etc.
list_concepts Enumerate every domain concept, optionally filtered by category.
get_pricing Public pricing, per-endpoint quotas, API-key location and plan recommendations.
call_readonly_endpoint Authenticated GET-only endpoint test runner. Blocks Proxy, Steam Guard and Tradeoffer actions.

Resources the MCP server exposes

  • docs://swagger Live OpenAPI 3.0 specification (5-minute cache).
  • docs://llms-full Curated long-form overview with all games, markets, products, endpoint guidance and competitor context.
  • /docs/raw/endpoints.md LLM-ready endpoint knowledge layer for every public OpenAPI operation.
  • /docs/raw/market-price-api.md Market price, item history and CS2 market index endpoint guide.
  • /docs/raw/inventory-api.md Inventory, profile and account-data endpoint guide.
  • /docs/raw/trade-api.md Trade offer, Steam login and Steam Guard endpoint guide.
  • /docs/raw/float-api.md CS2 float, inspect-link and screenshot endpoint guide.
  • docs://markdown/<file> Static documentation files (price-doc, response-format, market-index history, …).
  • docs://markdown/phases.json CS2 Doppler phase image-hash table.

How it works

Your AI agent talks to the MCP server over Streamable HTTP. The server reads the live OpenAPI specification of steamwebapi.com so every endpoint, parameter and response stays in sync with the production API. Domain concepts (Doppler phases, wear ranges, pricing fields, assetid vs classid) are exposed as a curated glossary.

Discovery tools work without an API key. The optional read-only executor accepts your existing steamwebapi API key via the MCP Authorization or X-Api-Key header and only calls documented GET endpoints. Mutating Tradeoffer, Steam Guard and Proxy actions are blocked by design. Use get_pricing when the agent should pick a package, check endpoint quotas or explain when to upgrade.