DCR Demo MCP
Demonstrates OAuth 2.0 Dynamic Client Registration with Duo SSO, providing three MCP servers for calendar events, document management, and analytics dashboards.
README
DCR Demo: MCP Servers + Duo SSO
A demo showing OAuth 2.0 Dynamic Client Registration (RFC 7591) with Duo SSO for MCP server authentication. Three MCP servers run over HTTP with full OAuth auth gates — both a chatbot web portal and Claude Code must authenticate via Duo before accessing any tools.
Architecture
┌──────────────────────────┐ ┌────────────────────────┐
│ Chatbot Portal (:8080) │ │ Claude Code / Codex │
│ (this web app) │ │ (MCP client) │
└──────────┬───────────────┘ └──────────┬─────────────┘
│ │
│ HTTP + Bearer token │ HTTP + Bearer token
│ │
┌──────┴──────┬──────────────┬────────────┴───────┐
│ │ │ │
▼ ▼ ▼ │
┌────────┐ ┌──────────┐ ┌───────────┐ │
│Calendar│ │ Documents│ │ Analytics │ │
│ :3001 │ │ :3002 │ │ :3003 │ │
└───┬────┘ └────┬─────┘ └─────┬─────┘ │
│ │ │ │
└─────────────┴──────────────┴─────────────────────┘
│
401 → RFC 9728
→ Discover Duo AS
→ DCR → PKCE → Token
│
┌────▼────┐
│ Duo SSO │
│ (AS) │
└─────────┘
Quick start
git clone https://github.com/cmedfisch/dcr-mcp-demo.git
cd dcr-mcp-demo
./setup.sh
Then configure your Duo SSO issuers:
cp config.json.example config.json
Edit config.json and replace the placeholder issuer URLs with your real Duo SSO URLs (one per server — see Configuration below).
Then start everything:
.venv/bin/python3 servers.py --all # MCP servers on :3001, :3002, :3003
.venv/bin/python3 app.py # Chatbot portal on :8080
Important: You must configure
config.jsonwith valid Duo SSO issuer URLs before starting the MCP servers. The servers read the config at startup and serve the issuer in their/.well-known/oauth-protected-resourceresponse. Without valid issuers, clients will attempt DCR againsthttps://example.comand fail. If you change the config after starting, restart the servers.
Configuration
Edit config.json to set the Duo SSO issuer for each server (one issuer per integration):
{
"calendar": {
"issuer": "https://sso-xxx.test.sso.duosecurity.com/oauth2/DIXXXXXXXXXX"
},
"documents": {
"issuer": "https://sso-xxx.test.sso.duosecurity.com/oauth2/DIXXXXXXXXXX"
},
"analytics": {
"issuer": "https://sso-xxx.test.sso.duosecurity.com/oauth2/DIXXXXXXXXXX"
}
}
Each server needs its own issuer URL (different IKEY) from a separate Duo SSO integration.
You can also configure issuers via the web UI at http://localhost:8080/config, but you must restart the MCP servers afterward for changes to take effect.
Add to Claude Code
Once the MCP servers are running (python3 servers.py --all):
claude mcp add dcr-calendar --transport http http://localhost:3001/mcp
claude mcp add dcr-documents --transport http http://localhost:3002/mcp
claude mcp add dcr-analytics --transport http http://localhost:3003/mcp
Claude Code will automatically:
- Hit the MCP server → get 401
- Discover the authorization server via
/.well-known/oauth-protected-resource(RFC 9728) - Register via DCR with Duo
- Open a browser for user authentication
- Exchange the code for a Bearer token
- Retry with the token → tools are now visible
Duo Admin Setup
This demo requires 3 SSO integrations (one per MCP server) and 3 custom agents registered in the Agent Directory. See the Duo MCP OAuth guide for full setup instructions.
Integrations (Duo Admin → Single Sign-On)
Create one SSO integration for each server. Each produces its own issuer URL (the IKEY at the end differs):
| Integration | Protects | Issuer pattern |
|---|---|---|
| Calendar MCP Server | :3001 |
https://sso-xxx.../oauth2/DI_CALENDAR_IKEY |
| Documents MCP Server | :3002 |
https://sso-xxx.../oauth2/DI_DOCUMENTS_IKEY |
| Analytics MCP Server | :3003 |
https://sso-xxx.../oauth2/DI_ANALYTICS_IKEY |
Agent Directory (Duo Admin → Agents)
Register 3 agents — one per server. The client_name sent during DCR must match the agent's DCR matching rule (EXACT or PARTIAL):
| Agent Name | DCR Match | Binds To |
|---|---|---|
| Calendar MCP Server | EXACT | Calendar integration |
| Documents MCP Server | EXACT | Documents integration |
| Analytics MCP Server | EXACT | Analytics integration |
Redirect URIs
Add these to each integration's allowed redirect URIs:
Chatbot portal (port 8080):
http://localhost:8080/callback/calendar
http://localhost:8080/callback/documents
http://localhost:8080/callback/analytics
Claude Code (MCP client SDK):
http://localhost/callback
http://127.0.0.1/callback
How DCR works
The client_name in the DCR registration payload is the agent's identity string. Duo admins configure DCR matching rules (EXACT or PARTIAL) that bind agents to Agent Classes based on this value.
Agent Duo SSO
| |
|--- POST /register --------------->| { "client_name": "Calendar MCP Server", ... }
|<-- 201 { client_id } -------------|
| |
|--- GET /authorize?client_id&pkce->| (user authenticates in browser)
|<-- redirect ?code ----------------|
| |
|--- POST /token (code+verifier) -->|
|<-- { access_token, id_token } ----|
| |
|=== Bearer token → MCP tools === |
The 3 MCP Servers
| Server | Port | Tools |
|---|---|---|
| Calendar | 3001 | list_events, get_event, create_event, check_availability |
| Documents | 3002 | list_documents, get_document, search_documents, upload_document, list_folders |
| Analytics | 3003 | get_metrics, get_audit_log, get_dashboard_summary, query_usage |
Files
servers.py— 3 HTTP MCP servers with OAuth auth gatesapp.py— Chatbot portal (Flask web app)config.json— Issuer configuration per serversetup.sh— One-liner setup scriptrequirements.txt— Python dependenciesmcp_server.py— Legacy stdio MCP server (kept for reference)
Stopping the servers
Use Ctrl+C in each terminal to stop cleanly. If ports are stuck:
# Find and kill processes on the demo ports
lsof -ti:3001,3002,3003,8080 | xargs kill -9
Or kill just the MCP servers:
lsof -ti:3001,3002,3003 | xargs kill -9
Requirements
- Python 3.10+
- A Duo SSO instance with DCR enabled
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.