Niagara MCP Server
Dynamically discovers and exposes Azure API Management APIs as MCP tools, enabling Copilot Studio to call them via natural language without hardcoding.
README
Niagara MCP Server — APIM Dynamic Discovery
An MCP (Model Context Protocol) server deployed on Azure that dynamically reads all APIs from Azure API Management and exposes them as AI tools to Copilot Studio. No hardcoding. No static files. APIM is the single source of truth.
Branch: Active development on
development.mainis protected.
How It Works
Copilot Studio (user types plain English)
|
| MCP protocol — Streamable HTTP
v
registry-mcp-server (Azure App Service — Docker container)
├── http_app_apim.py → exposes /mcp endpoint, accepts Copilot Studio connections
├── server_apim.py → MCP protocol: handles tools/list and tools/call
├── apim_dynamic.py → reads all APIs live from APIM Management API, filters, routes
├── registry.py → fallback only (returns empty if tools_registry.json missing)
├── dispatcher.py → fallback HTTP dispatcher for registry-based tools
└── apim_poc/policy.xml → APIM native gateway policy (POC reference, not active)
|
| Real HTTP calls with subscription key
v
Azure API Management (apim-dev-intg)
├── Function Apps (fn-dev-bark, fn-dev-dimitri-getshipmentfromwms, fn-dev-purchase-orders ...)
├── Logic Apps (dimitri-otmtodimitriupdate, ebs-to-dimitri, service-cloud-integration ...)
└── REST APIs (livetracking-getShipments, uomconversiontoerp, otmgenericapi ...)
Code Files
| File | What it does |
|---|---|
app/mcp_server/apim_dynamic.py |
Core. Authenticates via Managed Identity, calls management.azure.com, reads all 56 APIs, filters out EDI/B2B prefixes, converts each operation to MCP tool format, routes tool calls to correct APIM endpoint |
app/mcp_server/server_apim.py |
MCP protocol layer. Handles tools/list (calls apim_dynamic) and tools/call (routes to apim_dynamic or fallback) |
app/mcp_server/http_app_apim.py |
Web server. Exposes /mcp Streamable HTTP endpoint for Copilot Studio and /sse legacy endpoint |
app/mcp_server/registry.py |
Fallback only. Loads tools from tools_registry.json if it exists. File is deleted — returns empty dict |
app/mcp_server/dispatcher.py |
Fallback HTTP dispatcher. Used only if APIM discovery fails |
apim_poc/policy.xml |
APIM native gateway approach (Plan A / POC). Documents rate limiting, JWT auth, and CORS policies |
API Discovery Logic (apim_dynamic.py)
- Authenticates using App Service Managed Identity — no passwords, no hardcoded credentials
- Calls
management.azure.com→ reads all 56 APIs fromapim-dev-intg - Checks for
mcp-enabledtag — if any API is tagged, shows only tagged ones. If none tagged, shows all - Filters out EDI/B2B system callbacks by prefix:
o2c_,p2p_,b2b_,echo_api,fourkites_,fap_dev_,3pl_ofc_,fn_dev_edi_ - Converts remaining 53 business tools to MCP format with name, description, and input schema from APIM metadata
- Returns to Copilot Studio — happens live on every session, no caching
Any new Logic App or Function App published to APIM automatically appears in Copilot Studio on the next session. Zero code change, zero redeploy.
Security Controls
| Control | Status |
|---|---|
| Managed Identity — no credentials in code | Active |
| All secrets in App Service env vars — never in code or git | Active |
| APIM audit logging via Azure Monitor | Active automatically |
| EDI/B2B filter — AI cannot see system integration endpoints | Active |
mcp-enabled tag — whitelist specific APIs for AI access |
Ready — APIM admin adds tags |
| Rate limiting — 100 calls/min per subscription | Ready — one line in policy.xml |
| JWT Azure AD authentication on /mcp endpoint | Ready — one line in policy.xml |
Tech Stack
| Technology | Version | Purpose |
|---|---|---|
| Python | 3.11 | Server language |
| MCP SDK | 1.28.1 | MCP protocol |
| Starlette | 0.41.3 | Web framework |
| Uvicorn | 0.32.1 | ASGI server |
| HTTPX | 0.28.1 | Async HTTP calls to APIM |
| Azure Identity SDK | ≥1.15.0 | Managed Identity authentication |
| Docker | — | Containerization |
| Azure App Service | — | Hosting |
| Azure Container Registry | acrregistrymcp | Docker image storage |
| APIM Management API | 2022-08-01 | Dynamic API discovery |
Azure Resources
Resource group: rg-dev-IntegrationTeamAI — Subscription: Niagara Development
| Resource | Type | Purpose |
|---|---|---|
registry-mcp-server |
App Service | MCP server — live at azurewebsites.net |
acrregistrymcp |
Container Registry | Docker image storage |
apim-dev-intg |
API Management | Source of all 56 APIs |
Environment Variables (App Service)
| Variable | Purpose |
|---|---|
AZURE_SUBSCRIPTION_ID |
Azure subscription hosting APIM |
APIM_RESOURCE_GROUP |
Resource group of APIM instance |
APIM_SERVICE_NAME |
APIM name — currently apim-dev-intg |
APIM_BASE_URL |
APIM gateway URL |
APIM_SUBSCRIPTION_KEY |
Subscription key for All APIs product — sent on every tool call |
APIM_MCP_TAG |
Tag filter — default mcp-enabled |
Live URLs
| URL | Purpose |
|---|---|
https://registry-mcp-server.azurewebsites.net/mcp |
MCP endpoint — Copilot Studio connects here |
https://registry-mcp-server.azurewebsites.net/sse |
Legacy SSE endpoint |
https://apim-dev-intg.azure-api.net |
APIM gateway |
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.