Enterprise AI Gateway & Connection Management
Total Connections
Approved
Total Requests
Total Cost
⚠️ Admin action required — review and approve or deny each request below.
No pending approvals
ℹ️ Two telemetry views — same data, different depth
This table shows a lightweight summary of the last 50 proxied calls linked to this gateway.
The Live Telemetry Dashboard (sidebar → LLM Gateway → Live Telemetry) is the full-featured view: span types, trace IDs, per-request prompt/response detail, execution chain explorer, and advanced filters. Both pull from the same /telemetry endpoint.
| Timestamp | Agent | Destination | Status | Latency | Tokens | Cost | Details |
|---|---|---|---|---|---|---|---|
| Loading telemetry... | |||||||
Loading agents...
Approve or deny usage of non-API AI services such as Microsoft 365 Copilot, Copilot Studio, and browser-based assistants like ChatGPT, Gemini, and Claude.
Loading service policies...
Microsoft Graph external connections discovered from your M365 tenant via the Graph API.
⚠️ Usage Tracking Scope
Graph Connectors push external data into the Microsoft Search / Semantic Index. When M365 Copilot or a Copilot Studio agent retrieves from this index, the retrieval happens inside Microsoft's infrastructure and is not directly interceptable by the gateway proxy.
What can be tracked: If an agent explicitly calls the
/v1.0/search/query Microsoft Graph Search API
and that call is routed through this gateway, full telemetry is captured.
Aggregate search usage (not per-agent) is also available in the M365 Search admin analytics portal.
Click Refresh to load Graph Connectors from M365…
Blank = Authorization: Bearer {key}
No authentication — the agent will supply credentials at call time, or the endpoint is public.
ℹ️ What Happens Next:
1. Request will be submitted for admin approval.
2. Once approved, you’ll receive an MCP connection string.
3. Use this string in your AI agent to route traffic through the MCP server.
4. All requests will be logged with comprehensive telemetry.
Your MCP connection string is ready. Copy this and use it in your AI agent configuration:
🚀 How to Use:
// In your AI agent code:
const mcpConfig = JSON.parse(YOUR_MCP_STRING);
// Make requests through MCP bridge
const response = await fetch(
mcpConfig.mcp_server + '/bridge/request',
{
method: 'POST',
headers: {'Content-Type': 'application/json'},
body: JSON.stringify({
agent_name: 'YourAgentName',
destination: mcpConfig.endpoint,
method: 'POST',
body: your_request_data
})
}
);Full record of this API or service call. Large bodies are truncated for display.