Estimated reading time: 5 minutes
New MCP Server Lets R&E Community Connect Their AI Agent of Choice to Ask Questions About the Internet2 Network
A university reports that network traffic to a particular destination is taking longer than it should, or that a route just doesn’t look right. Before anyone touches a router, there’s a normal sequence of steps: check the monitoring dashboard, glance at the BGP Route Report to see whether the advertised routes match what’s expected, and rule out the easy explanations. Only once those first checks raise more questions than they answer does an engineer open Internet2’s Looking Glass and start running commands, traceroutes, route tables, and BGP summaries across several production devices at once to see what a given route actually looks like across the network.
That step has traditionally been where things slowed down. It’s manual work, and it assumes years of command-line experience and deep familiarity with the network.
To make that process faster and more accessible, Internet2 has released a public MCP server that lets AI agents ask questions about the Internet2 network. It’s the first AI-enabled capability available through Periscope, the successor to Insight Console currently in development.
The server uses Model Context Protocol, the emerging standard for connecting AI agents to real tools and real data instead of just answering from training data. The capability available through Periscope’s MCP server today is Looking Glass. A custom internal tool, a coding assistant, or a chatbot your team already uses can be pointed at the server and start asking the Internet2 network direct questions, getting direct answers back. It’s the same data available through the Looking Glass interface at console.internet2.edu, without needing to know the right command syntax or device names.
The result is AI-assisted operations that lower the barrier to accessing network insights and help research and higher education (R&E) IT teams troubleshoot faster.
Periscope Builds AI Access In From the Start
Insight Console is the single web-based place the R&E community uses today to visualize, manage, and troubleshoot Internet2 network services, from Looking Glass to Routing Intentions and the BGP Route Report to Virtual Networks for multicloud connectivity. Periscope is being designed as the platform that carries that work forward, with AI-native access built in from the start. The MCP server and the Looking Glass access it offers give the community its first hands-on piece of that vision. Development continues from here, with more capabilities rolling out and the fuller platform on track to debut at the 2026 Internet2 Technology Exchange (TechEX26), Oct. 26-30 in Minneapolis.
The MCP server is also part of a broader push across Internet2 Network Services to make our tools AI-aware and AI-forward for the engineers who rely on them every day, advancing a key objective of the Trusted Infrastructure Program. Since October 2025, our AI Ops team has been demonstrating a working troubleshooting assistant built on this same foundation. That work has now expanded through an agent-to-agent interoperability proof of concept, demonstrating with our colleagues at SURF how AI agents can support troubleshooting across multiple network domains. This is likely the first of many cooperative efforts to come between Internet2, regional and state R&E networks, and national R&E networks as we work toward more seamless, efficient troubleshooting globally.
The ticket, the dashboard check, and the route report review still come first. What changes is the step after that. Command-line fluency in Looking Glass used to be a prerequisite; now more people on the team can access the same network insights, while experienced users can identify and fix root causes faster.
When to Reach for the MCP Server
A few rules of thumb for using it well:
- For quick questions: Ask conversational questions when you’d otherwise open Looking Glass just to double-check something, like whether anything unusual is happening on the Chicago routers right now. Talk to it the way you’d ask a colleague.
- For when AI is already part of the workflow: If your team already runs its own AI agent for other work, you can connect it to Periscope. That means network questions can come up naturally in the middle of that work, instead of switching over to a browser tab.
- For less frequent Looking Glass users: Anyone who doesn’t run Looking Glass commands every day benefits, too. A new team member can ask in plain language rather than looking up the right command and device name first.
The console still has its place. Making comparisons across many devices, saving and sharing a specific command’s output via URL, or simply wanting the full Looking Glass interface in front of you: those are reasons to go to console.internet2.edu directly. For engineers well-versed in using Looking Glass, it will continue to be the right choice for in-depth, hands-on exploration. As Periscope continues in development, we see the console getting faster and having more tools available, as well.
Connect Your AI Agent and Start Exploring
Getting started is straightforward: point your MCP-compatible AI agent at Periscope’s MCP server (https://periscope.ns.internet2.edu/mcp) and start asking questions. Full setup instructions are available in the Console Documentation.
If you run into questions or want a hand getting connected, the Internet2 Network Operations Center is the place to start.
Join Us at Technology Exchange to Learn More