Developers live in their editors. SREs live in their dashboards. When something breaks, the developer has to leave VS Code, open a browser, navigate to the APM tool, find the right service, locate the right time window, parse the trace waterfall, copy the stack trace, go back to VS Code, and figure out the fix.

That workflow is broken. Every tab switch is a context loss. Every context loss adds minutes to resolution.

What if your editor could talk directly to your APM?

That's what the Model Context Protocol makes possible. And we built an open-source MCP server that does exactly this for IAPM.

What Is MCP?

The Model Context Protocol is an open standard that lets AI assistants connect to external tools and data sources. Think of it as a universal adapter: any AI assistant that speaks MCP can talk to any server that implements it.

GitHub Copilot in VS Code supports MCP in agent mode. Claude Code supports it. Cursor supports it. The protocol is gaining adoption fast because it solves a real problem: AI assistants are smart, but they can only work with the context they can see. MCP lets them see your tools.

An MCP server for observability means your AI assistant can query your APM data directly. No browser. No dashboard. No copy-paste. You ask a question in your editor, and the assistant pulls the answer from your production telemetry.

What Our MCP Server Does

The IAPM MCP Server is open-source (MIT license), self-contained, and runs on Windows, macOS, and Linux. It connects your AI assistant to your IAPM grids, giving it access to your traces, errors, metrics, and service topology.

Once configured, you can ask questions like:

• "Were there any errors yesterday?"
• "Show me SQL errors in the last hour"
• "Show me the stack trace for that error"
• "How do I fix it?"

The assistant doesn't just search logs. It queries your actual telemetry data through IAPM's API, retrieves the relevant traces and spans, and presents the information in your editor. Then, because it's in your IDE and has access to your codebase, it can suggest fixes in the same conversation.

That last part is the key. Dashboard-based APM tools can show you the error. Your editor-based AI assistant can show you the error AND help you fix it. The MCP server bridges the gap.

Setup in Two Minutes

Download the latest release for your platform (Windows x64, macOS ARM64, Linux x64, Linux ARM64). The binaries are self-contained, no .NET runtime required.

Add this to your .vscode/mcp.json:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "if-key",
      "description": "Immersive APM API Key",
      "password": true
    }
  ],
  "servers": {
    "IF.APM.App.MCP.Server": {
      "type": "stdio",
      "command": "/path/to/IF.APM.App.MCP.Server",
      "env": {
        "IF_ApiKey": "${input:if-key}"
      }
    }
  }
}

Replace the command path with wherever you extracted the binary. VS Code prompts for your API key on first use. That's it.

Open GitHub Copilot Chat in agent mode and start asking questions about your application's performance. The MCP server handles the translation between natural language and IAPM's telemetry API.

Why Open Source Matters Here

We could have built this as a closed plugin. Ship a VS Code extension, control the distribution, gate it behind a subscription. That's what most vendors do.

We open-sourced it instead, under the MIT license, because the MCP ecosystem benefits from openness. The more MCP servers exist for observability tools, the more valuable the protocol becomes for everyone. Developers who use Datadog, Grafana, or New Relic should have MCP servers too. The protocol works best when it's universal.

Open-sourcing also means you can inspect exactly what the server does. It connects to IAPM's API using your API key. It translates MCP tool calls into telemetry queries. It returns the results. No telemetry collection. No usage tracking. No phone-home behavior. Read the code yourself: it's a .NET application, straightforward to audit.

The Workflow That Changes Everything

Here's what debugging looks like with the MCP server:

Before: Alert fires. Open browser. Navigate to IAPM. Find the service. Find the time range. Click through traces. Find the error. Copy the stack trace. Switch to VS Code. Paste. Read. Correlate with code. Fix. Test. Deploy. Total context switches: at least 6.

After: Alert fires. Open Copilot Chat. Type "What errors happened in the last 30 minutes?" The assistant queries IAPM via MCP, shows you the errors with stack traces, identifies the relevant code in your repo, and suggests a fix. Total context switches: zero.

The time savings compound. A single incident investigation that involved 10 minutes of tab-switching now takes 2 minutes of conversation. Across a team of engineers handling multiple incidents per week, that's hours recovered.

But the bigger win isn't time. It's accessibility. Junior engineers who don't know how to navigate a complex APM dashboard can ask questions in plain English. "Which service had the most errors this week?" "What changed in the last deployment?" "Why is the checkout service slow?" The MCP server democratizes access to telemetry data that used to require specialized knowledge to query.

Beyond VS Code

The MCP server works with any MCP-compatible client, not just VS Code. As the protocol matures and more AI assistants adopt it, the same server binary works everywhere. Claude Code, Cursor, Windsurf, and any future tool that implements the MCP client specification can connect to your IAPM data using the same server.

This is also why we built Tessa, our built-in AI Assistant, with the same philosophy. Tessa lives inside IAPM's 3D environment and has full access to your telemetry, topology, and codebase. The MCP server extends that same data access to your IDE. Two interfaces to the same observability intelligence: one immersive, one embedded.

Try It

1. Start Free and grab your API key
2. Download the MCP server binary for your platform
3. Add the config to .vscode/mcp.json
4. Open Copilot Chat in agent mode and ask about your application's health

Or, if you want to see the full 3D experience alongside the IDE integration, run our trace generator to populate your grid with realistic telemetry:

go install github.com/ImmersiveFusion/if-opentelemetry-tracegen/cmd/tracegen@latest
tracegen -endpoint otlp.iapm.app:443 -headers "api-key=YOUR_KEY" -complexity light

Then query that data from VS Code. Ask "What services are running?" and watch the MCP server return your full topology without opening a browser.

Your APM data belongs where you work. Not in a separate tab.

Start Free. Immersive. AI-guided. Full-stack observability. Enter the World of Your Application®.