Agile Analytics · Local AI connector

Azure DevOps sprint metrics for your own AI assistant

Ask GitHub Copilot in VS Code, Claude or another assistant you already use about your team's sprint, cycle time or forecast, and get numbers worked out by Agile Analytics instead of a guess. The local AI connector runs on your own computer, only reads, and sends nothing to Baytek.

What your assistant can ask for

The connector is an MCP server. MCP, the Model Context Protocol, is the standard way an AI assistant calls a tool on your computer: you ask in plain words, your assistant asks the connector, and the connector answers with numbers. It answers six questions.

Sprint summary

What a finished sprint delivered, what was committed at the end of day one, how much of it got done, and what carried over.

Cycle time

How long items took from start to done over the last 3, 6 or 12 finished sprints, with the ids of the slowest ones.

Flow efficiency

How much of that time items were actively worked on, and how much they spent waiting.

Work in progress

What is in progress in the running sprint right now, by workflow stage.

Aging work

Open items in the current sprint and how many days each one has been in its stage.

Monte Carlo forecast

How many items the team is likely to finish in a number of days, or when a backlog of a given size is likely to be done.

A status check tells you in one answer what is still missing before the first question. On Premium and during the trial, your assistant also gets three ready-made commands (sprint review, stand-up and forecast) and a link in every answer that opens the matching view in Agile Analytics.

An MCP server that counts the way your team does

The connector follows the workflow mapping your team saved in Agile Analytics, so “active” and “done” mean the same thing in your assistant as they do on your dashboards. Your assistant gets the result, not a pile of work items to count itself.

Microsoft's own Azure DevOps MCP server is a different tool: it gives an assistant raw work items to read, and it does not compute flow metrics.

For the longer explanation, read An MCP server for Azure DevOps flow metrics.

Three example conversations

The numbers below are made up, to show the kind of answer you get. Your assistant answers with your team's real numbers.

You type

“How did team Platform’s last sprint go? Project Web Store, organisation contoso.”

Sprint summary · example numbers

Your assistant tells you the team delivered 18 items worth 41 points, committed to 20 items at the end of day one, finished 16 of those, and carried 3 items into the next sprint. You can then ask it to draft notes for the sprint review from those numbers.

You type

“What has our cycle time been over the last 6 sprints, and which items took longest?”

Cycle time · example numbers

Your assistant tells you that half of the finished items took 4 days or less and 85% took 9 days or less, and lists the ids of the slowest items, such as #4127 and #4133, so you can open them in Azure DevOps.

You type

“How many items can team Platform finish in the next 30 days?”

Monte Carlo forecast · example numbers

Your assistant tells you that, based on the team’s throughput in recent finished sprints, there is an 85% chance of finishing at least 22 items and a 50% chance of finishing at least 27.

Is my data safe?

It runs on your own computer

The connector runs on your machine and signs in to Azure DevOps as you, with your Azure CLI sign-in or a read-only token. It sees only what you can already see in Azure DevOps.

It only reads

It cannot create, change or delete anything in Azure DevOps.

It sends nothing to Baytek

No telemetry, no usage counts and no copy of your data reach Baytek, the company behind Agile Analytics.

Numbers, dates and ids only

Answers hold numbers, dates and work item ids. Never a work item title, a person’s name or a comment.

Your AI provider sees the numbers

The answers become part of your conversation with your assistant, so they go to the AI provider you chose, like anything else you type there. Check that provider’s terms as you would for any prompt.

Your admin stays in control

An organisation admin can switch the connector off for everyone in the organisation.

What you need

Agile Analytics in your Azure DevOps organisation

Install the extension from the Visual Studio Marketplace. The free 30-day trial includes the connector.

A saved workflow mapping

Save the workflow mapping for the project you will ask about. The connector never guesses which states count as active or done.

An AI assistant that supports MCP

For example GitHub Copilot in VS Code or Claude, plus Node.js on your computer to start the connector. The set-up guide shows each step.

A seat

The connector works during the trial and on every paid plan, for people who hold a seat.

What it does not do

Six questions, not free-form queries

Your assistant can ask the six questions above. It cannot run its own queries over your Azure DevOps data.

Ids, not titles

Answers name work items by id. Open an item in Azure DevOps to see what it is.

One team at a time

Each question is about one team. Name the organisation, project and team when you ask.

FAQ

Can I ask GitHub Copilot or Claude about my Azure DevOps sprint?

Yes. With Agile Analytics installed and the local AI connector added to your assistant, GitHub Copilot in VS Code, Claude or another assistant that supports MCP can ask for your sprint summary, cycle time, flow efficiency, work in progress, aging work and a Monte Carlo forecast.

Does my Azure DevOps data go to Baytek?

No. The connector runs on your own computer, reads Azure DevOps as you and sends nothing to Baytek. The numbers it returns do go to the AI provider you chose, as part of your conversation with it.

Will the numbers match our Agile Analytics dashboards?

The connector follows the workflow mapping your team saved in the extension, so your assistant counts with the same definitions of active and done that your dashboards use.

Can the connector change our work items?

No. It is read-only. It cannot create, change or delete anything in Azure DevOps.

Which plans include it?

It works during the free 30-day trial and on every paid plan, for people who hold a seat. An organisation admin can switch it off for everyone.

Try it with your own team

Install Agile Analytics, save your workflow mapping, then follow the set-up guide to add the connector to your assistant.

Want the background first? Read how the MCP server works.