Keypup MCP Server: Bring Engineering Analytics to Any AI Assistant

Introduction: The Rise of the Model Context Protocol

Imagine asking Claude, ChatGPT, or your favorite AI assistant: "What was our average PR cycle time last month?" and getting an instant, accurate answer pulled directly from your unified Git and Jira data. No dashboard to build. No SQL to write. No switching contexts.

This is the promise of Model Context Protocol (MCP)—and it's rapidly becoming the standard for how AI assistants interact with enterprise data and tools. Introduced by Anthropic in late 2024, MCP has exploded in adoption, with major AI platforms like OpenAI, Google, and Mistral AI embracing the protocol. The reason is simple: MCP solves the "last mile" problem of AI—giving assistants secure, standardized access to your company's structured data.

For engineering teams, this represents a fundamental shift. Instead of logging into yet another analytics platform, you can simply ask your questions where you already work—in your code editor, in your AI chat window, or even in your terminal. The data comes to you, contextualized by an AI that understands both your question and your domain.

Today, we're thrilled to announce that Keypup is now MCP-enabled. Our new MCP server brings the full power of Keypup's unified engineering analytics directly into any MCP-compatible AI client, letting you query SDLC metrics, drill into bottlenecks, and explore your data using plain English.

Context: Why MCP Servers Are Transforming Engineering Workflows

The software development landscape is drowning in tools. The average engineering team uses 10+ platforms: GitHub or GitLab for code, Jira or Linear for project management, Slack for communication, and specialized analytics tools like Keypup for understanding performance. Each tool has its own interface, its own query language, its own way of presenting data.

This fragmentation creates friction. When a VP of Engineering wants to understand why delivery is slowing down, they need to:

1. Log into their analytics platform
2. Navigate to the right dashboard (or build a new one)
3. Configure the right filters and timeframes
4. Export or screenshot the results
5. Context-switch back to where they were working

By the time they have their answer, they've lost 10 minutes and broken their flow.

This is where Model Context Protocol shines. MCP creates a standardized bridge between AI assistants and structured data sources. Instead of building custom integrations for every AI platform, tool providers like Keypup can implement one MCP server that works with every MCP-compatible client.

For engineering leaders, the value proposition is immediate:

• Context preservation: Ask questions without leaving your workflow
• Natural language interface: No query languages or dashboard configuration
• AI orchestration: The assistant handles the complexity of multi-step queries
• Iterative refinement: Follow-up questions build on previous context

The MCP trend is accelerating rapidly. In Q1 2026 alone, over 50 major developer tools announced MCP support, and GitHub reported that 35% of Copilot Enterprise users are now using MCP-connected tools. The protocol is becoming as fundamental to AI-powered development as REST APIs were to the web.

Keypup's MCP server positions your engineering team at the forefront of this shift.

What the Keypup MCP Server Can Do

The Keypup MCP Server exposes a focused toolkit that lets AI assistants explore and query your engineering data autonomously. Think of it as giving Claude, ChatGPT, or Cursor direct access to your Keypup analytics—but with the AI handling all the complexity.

The Five Core Tools

When you connect the Keypup MCP Server to your AI client, five powerful tools become available:

1. Discover Your Companies Lists the Keypup companies (teams/organizations) you belong to. This is always the first step—queries need to know which company's data to analyze.

2. Explore Datasets Browse the available datasets: issues & pull requests, commits, comments, reviews, and activity events. Each dataset represents a different facet of your development lifecycle, and the AI can examine all of them to determine the best data source for your question.

3. Discover Formula Operators Lists the functions and operators available when building metrics, dimensions, and filters. This gives the AI a complete vocabulary for constructing precise queries—from date functions like YEAR_MONTH() to aggregators like COUNT() and AVG().

4. Explore Dataset Fields For any chosen dataset, this tool reveals all available fields (columns), their types (string, number, date, array), and their descriptions. The AI can search for specific fields by pattern, making it easy to find exactly what it needs—even custom Jira fields unique to your organization.

5. Run Reporting Queries Executes structured queries with metrics (aggregations), dimensions (group-by), filters, sorting, and pagination. This is where the AI actually fetches your data and brings back results.

The Secret Weapon: AI-Powered Query Generation

Here's where it gets powerful. In addition to the five core tools, Keypup's MCP server includes a query generation tool that leverages AI to translate your natural language request into a ready-to-run structured query.

This means the AI assistant doesn't just have access to your data—it has access to another AI that specializes in building Keypup queries. It's AI orchestrating AI, and the result is remarkably sophisticated query construction without any manual intervention.

How a Query Actually Works

Let's trace what happens when you ask: "Show me our PR merge rate by team over the last quarter."

{
  "dataset": "ISSUES_PULL_REQUESTS",
  "metrics": [
    { "label": "Merged PRs", "formula": "COUNT_IF(state == 'MERGED')" }
  ],
  "dimensions": [
    { "label": "Week", "formula": "YEAR_WEEK(merged_at)", "sort": "DESC" },
    { "label": "Team", "formula": "team" }
  ],
  "filters": [
    { "formula": "type == 'pull_request'" },
    { "formula": "merged_at >= NOW() - 3 * MONTH()" }
  ]
}

All of this happens in seconds. From your perspective, you asked a question and got an answer. But under the hood, the AI orchestrated 5-6 tool calls, built a sophisticated query with proper filters and aggregations, and formatted the results in a human-readable way.

This is the power of MCP: giving AI assistants the tools they need to be truly autonomous agents over your data.

Real-World Questions You Can Ask

Once the Keypup MCP Server is connected, your AI assistant becomes a conversational interface to your entire engineering analytics platform. Here are real examples of questions teams are asking:

Delivery & Throughput Insights

Other common throughput questions:

• "What's our weekly issue closing rate this quarter?"
• "How many commits were made per author last month?"
• "Show me story points completed by sprint over the last 6 sprints."

Cycle Time & Performance Analysis

Other performance questions:

• "Which repositories have the slowest cycle time?"
• "Show me review turnaround time trend for the last 3 months."
• "What's the correlation between PR size and time to merge?"

Quality & Process Questions

Other quality questions:

• "What proportion of our pull requests resolve at least one issue?"
• "How many PRs were merged without a review?"
• "Show me our change failure rate by week."

Workload & Collaboration Insights

Other collaboration questions:

• "How is work distributed across the team right now?"
• "How many comments do our pull requests receive on average?"
• "Which PRs have been waiting for review the longest?"

Open-Ended Exploration

The beauty of natural language is that you don't need to know the exact query to ask. You can explore broadly:

• "Summarize our engineering activity over the last month."
• "Compare open vs. closed issues over time."
• "What labels are most common on our issues?"
• "Are there any unusual patterns in our commit activity?"

And because the conversation is iterative, you can refine:

• "Now break that down by repository."
• "Restrict it to the backend team."
• "Show me only critical bugs."

The AI maintains context across the conversation, making data exploration feel natural rather than mechanical.

Setting Up the Keypup MCP Server in 3 Steps

The entire setup process takes less than 5 minutes. Here's how:

Step 1: Generate Your Keypup API Token

1. Log into Keypup and navigate to Settings → Account → API Token
2. Click Create token
3. Set a descriptive name (e.g., "MCP - Claude Desktop")
4. Important: Set an expiration date (we strongly recommend this for security)
5. Select token scopes:
   • ✅ Reports read (required to query datasets)
   • ✅ Companies read (required to list your companies)
6. Click Create and immediately copy the token (it's only shown once)

Step 2: Configure Your MCP Client

Most MCP clients (Claude Desktop, Cursor, VSCode, Kiro) use a JSON configuration file. Here's how to add Keypup:

For Claude Desktop (on macOS: ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "keypup": {
      "url": "https://hq.keypup.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN_HERE"
      }
    }
  }
}

For Cursor (settings.json):

{
  "mcp.servers": {
    "keypup": {
      "url": "https://hq.keypup.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN_HERE"
      }
    }
  }
}

For VSCode (with MCP extension):

{
  "mcp.servers": [
    {
      "name": "keypup",
      "url": "https://hq.keypup.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN_HERE"
      }
    }
  ]
}

Replace YOUR_API_TOKEN_HERE with the token you generated in Step 1. Keep the Bearer prefix.

Step 3: Test the Connection

Restart or reconnect your MCP client, then try a simple test query:

If the AI successfully returns your company names, you're connected! Try a more complex query:

The Benefits: Why This Changes Everything

1. Zero Context Switching

Engineering leaders and developers can stay in their flow. Writing code in Cursor? Ask about PR review metrics without opening a browser. Having a strategy discussion in Claude? Pull in real SDLC data to support your argument. The data comes to you.

2. No Query Language Expertise Required

You don't need to know SQL, GraphQL, or any proprietary query syntax. Natural language is the interface. This democratizes data access—junior developers can ask the same sophisticated questions as seasoned data analysts.

3. AI Handles the Complexity

The Keypup MCP Server doesn't just fetch data—it understands your schema. It knows which dataset to use for which question. It knows how to construct proper filters and aggregations. It even handles pagination and sorting automatically. You get all the power of Keypup's analytics engine without needing to understand how it works.

4. Iterative and Contextual

Unlike traditional dashboards where each query is isolated, MCP-enabled conversations are contextual. Ask a question, get an answer, then refine: "now show me just the backend team" or "break that down by repository." The AI maintains state across the conversation.

5. Works Everywhere

Because MCP is an open standard, Keypup's MCP Server works with any MCP-compatible client. Today that's Claude, Cursor, ChatGPT, VSCode, Gemini, Mistral, and Kiro. Tomorrow it could be a new AI assistant you haven't even heard of yet. You set up once and benefit everywhere.

6. Enterprise-Grade Security

Despite feeling conversational, the MCP Server maintains enterprise security standards:

• Scoped permissions: Your API token only grants the access you explicitly allow
• Read-only by default: MCP tools never modify your data
• Token expiration: Time-limited access with periodic rotation
• Data isolation: Queries are automatically scoped to your authorized companies

You get the ease of natural language with none of the security compromises.

The Bigger Picture: MCP and the Future of Engineering Intelligence

The Keypup MCP Server is part of a larger shift in how engineering teams interact with their data. We believe the future of engineering intelligence isn't about better dashboards—it's about ambient intelligence: data that's always available, always contextual, and always conversational.

Imagine a near future where:

• Your IDE notices you're working on a slow-performing service and surfaces relevant performance metrics without you asking
• Your standup meeting automatically includes a verbal summary of yesterday's engineering activity
• Your quarterly planning sessions have an AI assistant that can instantly answer "what if" questions about capacity and velocity

This isn't science fiction. With MCP, the building blocks are already here. And Keypup is positioned at the intersection of unified SDLC data and conversational AI—the exact combination needed to make this future real.

Getting Started Today

The Keypup MCP Server is available now in public beta for all Keypup customers. We're actively expanding the capabilities during this phase and welcome your feedback.

Ready to bring your engineering analytics to your AI assistant?

1. Generate your API token: Keypup Settings → API Token
2. Read the full documentation: Using the MCP Server
3. Join our community: Share your use cases and provide feedback via our in-app chat

If you're not yet a Keypup customer but want to experience the power of conversational engineering analytics, start your free trial today.

The era of dashboard-driven engineering intelligence is ending. The era of conversational, ambient intelligence is beginning—and with Keypup's MCP Server, you can be part of it from day one.