APS MCP Auth Examples
Reference MCP servers that integrate with Autodesk Platform Services, demonstrating various authentication combinations and exposing tools to list projects and folder contents via the Data Management API.
README
APS MCP Auth Examples
Reference implementations of MCP servers that integrate with Autodesk Platform Services, covering every combination of:
- How the server authenticates to APS
2LO- 2-legged OAuth, app-wide3LO- 3-legged OAuth, per-userPKCE- 3-legged OAuth for public clients, per-userSSA- Secure Service Account, non-human identity
- How MCP clients authenticate to the server
- not at all (STDIO)
- via an external identity provider
- via an OAuth proxy service
All HTTP-based examples target the 2026-07-28 MCP specification revision
via the split MCP TypeScript SDK v2
(@modelcontextprotocol/{server,express,node}), which requires MCP servers to
act as OAuth 2.1 resource servers backed by a dedicated authorization server,
and prefers Client ID Metadata Documents (CIMD) over Dynamic Client
Registration for identifying MCP clients.
Every example exposes the same MCP tools, implemented once in shared/ and
reused everywhere: list-projects (hubs + projects, via the Data Management
API) and list-contents (a project's top-level folders, or a specific folder's
contents).
The examples
| Folder | MCP-client auth | What it demonstrates |
|---|---|---|
aps-mcp-server-local |
none (STDIO) | The simplest possible setup — a locally spawned process, no MCP-layer auth at all. |
aps-mcp-server-remote-auth0 |
External IdP (Auth0) | This server only verifies tokens; Auth0 (or any OIDC/JWKS provider) remains the authorization server. Per-IdP-user APS providers cached in memory. |
aps-mcp-server-remote-proxy |
Separate OAuth proxy service | Relies on an OAuth proxy in front of APS authentication (simple-oauth-proxy) to generate "MCP tokens", and uses /internal/exchange endpoint to exchange these for "APS tokens". |
simple-oauth-proxy |
(is the proxy) | The standalone, provider-agnostic OAuth proxy service consumed by aps-mcp-server-remote-proxy, built with Python + FastMCP. |
shared |
(library) | The five APS auth provider classes, the two MCP tools, and small helpers reused by every example above. |
Setup common to every example
- Register an APS application at https://aps.autodesk.com/myapps (a
Traditional Web App if you'll use any
3LOexample; a Server-to-Server / API-key style app is enough for2LO-only use). ForSSA, additionally create a Secure Service Account and register its public key — see the SSA guide. npm installat the repo root — this is an npm workspaces project, so one install resolvessharedand all four TypeScript servers.- Run any TypeScript example with
npm startfrom inside its folder (ornpm run start -w <package-name>from the root), after copying its.env.exampleto.envand filling in the values for your chosenAPS_AUTH_MODE. simple-oauth-proxyis a separate Python/FastMCP service — see its own README for setup; it's only needed if you're tryingaps-mcp-server-remote-proxy.
A note on scope
These are teaching examples, optimized to be read end-to-end in one sitting. Several corners intentionally cut for brevity are called out in the relevant README (in-memory-only state with no horizontal-scaling story, a simplified CIMD fetch without full SSRF hardening in the OAuth proxy example, etc.). Don't copy the security-relevant bits verbatim into production without reading those notes.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。