Rally has an MCP Server now. As in, your AI assistant can talk to Rally without you opening a browser tab. We've been waiting for this one.

MCP, Model Context Protocol, is the open standard that lets AI tools like Claude Code and Cursor connect to external systems. Rally's implementation means you can query work items, create stories, update defects, and manage tasks from inside your IDE. No copy-pasting between windows, no "let me go check Rally" interruptions in the middle of writing code.

What It Actually Does

The MCP Server exposes Rally's WSAPI through a set of tools that any MCP-compatible client can call. In practical terms:

  • Read and query artifacts, pull up your current sprint's stories, check acceptance criteria, find that defect somebody mentioned in standup
  • Create and update work items, log a defect from your IDE when you find one, update task status without switching windows

Permissions are scoped through API keys, so your AI agent gets only the access you give it. This matters when your security team asks questions, and they will.

It covers what Broadcom calls "the most common workflows for Rally users." It's not the full WSAPI surface, don't expect to reconfigure your subscription through it, but for day-to-day development work, it handles the stuff you actually need.

How to Turn It On

This isn't a checkbox in your Rally settings. Your Subscription Admin needs to open a support ticket with Broadcom to enable MCP Server for your subscription. Once it's on, you configure your IDE's MCP client to point at it with an API key.

Broadcom has documentation for Cursor specifically, and other MCP-compatible clients should work with similar configuration. If your team uses Claude Code, the setup is straightforward, point it at Rally's MCP endpoint and authenticate.

Why This Is a Big Deal for Rally Teams

Rally has been browser-only for its entire life. Every developer interaction with Rally meant leaving your IDE, navigating to the right project, finding the right artifact, making your update, and going back to your code. Multiply that by every developer on your team, every day, and you get a meaningful chunk of lost focus.

MCP changes that. Your AI assistant becomes a bridge to Rally, and because MCP is a protocol, not a proprietary plugin, it works across different AI tools. You're not locked into one vendor's IDE integration.

For enterprise teams running CI/CD pipelines, the automation angle is worth paying attention to. AI agents can update Rally as part of your build and deploy workflow, mark stories as delivered, attach build information, flag failing tests against the relevant defect. The kind of bookkeeping that everyone agrees should happen but nobody wants to do manually.

We've Been Using This

Full disclosure: we use MCP every day. Not for Rally specifically, our whole workflow runs through Claude Code with MCP connections to various systems. So when Rally shipped their MCP Server, we were already familiar with the protocol and what it can do. That's not us bragging. That's us admitting we spend our days talking to AI assistants about code, which is either impressive or concerning depending on who you ask.

The point is, if you have questions about how MCP works in practice, not just what the docs say, but what it's like to actually build workflows around it, we have opinions. Informed ones, mostly.

Want Help Setting This Up?

If you're an enterprise Rally team thinking about MCP and you're not sure where to start, how to scope permissions, which workflows to automate first, how to roll it out without your security team losing sleep, we should talk.