Local AI connector
Ask Claude or another AI assistant on your own computer about your sprints and flow, and get the same numbers the Agile Analytics views show. The connector reads only counts, dates and work item ids, with your own Azure DevOps access, and sends nothing to Baytek.
What it is
An optional connector, part of your Agile Analytics subscription, that you install on your own computer with npx. It lets Claude or another assistant that can run a local MCP server read Agile Analytics numbers with your own Azure DevOps access.
Scrum Masters, team leads and managers who already ask an AI assistant to help with sprint reviews, retros and planning, and want it to use the same numbers the Agile Analytics views show instead of guessing.
Agile Analytics installed in your Azure DevOps organisation, a saved workflow mapping for the project, Node.js on your computer (npx comes with it), and an assistant that supports local MCP servers.
Install
Add the connector to your assistant's settings. npx downloads it the first time it starts, so there is nothing else to install.
Claude Desktop
Settings → Developer → Edit Config opens claude_desktop_config.json. Add this, save, and restart Claude Desktop.
{
"mcpServers": {
"agile-analytics": {
"command": "npx",
"args": ["-y", "@baytekdev/agile-analytics-mcp"]
}
}
}Claude Code
Put this in .mcp.json at the root of a project, or run the one-line command below instead.
{
"mcpServers": {
"agile-analytics": {
"command": "npx",
"args": ["-y", "@baytekdev/agile-analytics-mcp"]
}
}
}One line: claude mcp add agile-analytics -- npx -y @baytekdev/agile-analytics-mcp
VS Code
Put this in .vscode/mcp.json in your workspace, or in your user mcp.json.
{
"servers": {
"agile-analytics": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@baytekdev/agile-analytics-mcp"]
}
}
}Sign in to Azure DevOps
The connector reads Azure DevOps as you, so it sees exactly what you can already see. It tries these in order.
In Azure DevOps, open User settings → Personal access tokens → New Token and give it exactly three scopes: Work Items (Read), Project and Team (Read) and Extension Data (Read). On Windows, set it as a user environment variable named AZURE_DEVOPS_PAT (setx AZURE_DEVOPS_PAT "your-token") and keep it out of the config file. On macOS and Linux, desktop assistants do not read your shell profile, so put it in the server entry's "env" block as AZURE_DEVOPS_PAT and keep that config file readable by your account only. Restart your assistant afterwards. The connector keeps the token in memory only and never prints, logs or writes it.
Works for organisations connected to Microsoft Entra ID; Microsoft-account-only organisations cannot use it. Leave AZURE_DEVOPS_PAT unset and run az login once. If you belong to several tenants, add "--tenant", "<tenant id>" to the args.
What you can ask
Six questions, each counted the way the matching view counts it. Name the team, the project and the organisation (the part after dev.azure.com/) in your question.
One finished sprint as Sprint Summary counts it: delivered items and points, cycle time, carryover, commitment, reliability (say/do) and scope added and removed.
“Sprint summary for team Platform in project Web Store, org contoso.”
Count, average and p50/p70/p85/p95 days over the last 3, 6 or 12 finished sprints, as the Cycle Time view plots it, with the ids of the slowest items.
“Cycle time over the last 6 sprints for team Platform in project Web Store, org contoso.”
Active time against waiting time over the last 3, 6 or 12 finished sprints, as the Flow Efficiency view works it out.
“Flow efficiency for team Platform in project Web Store, org contoso.”
Items in progress right now in the running sprint, by workflow stage, against your saved WIP limit.
“How much work is in progress for team Platform in project Web Store, org contoso?”
Open work in the current sprint, carried-over items included: each item’s id, stage, days in that stage and days in flow, on the same weekend basis as your Aging view.
“Which items are aging for team Platform in project Web Store, org contoso?”
A Monte Carlo forecast from the last 12 sprints’ throughput: how many items in a number of days, or when a backlog of a given size finishes.
“How many items can team Platform in project Web Store, org contoso, finish in the next 30 days?”
What it reads, and what it never reads
State, type, dates, sprint and area path, estimates, board column and the blocked flag, and their state and sprint history.
The team’s sprints and working days, and the area paths of the project’s teams, to count the same items the views count.
The saved workflow mapping, Item Types, your WIP limit and weekend choice, and the admin switch. It never guesses a mapping.
dev.azure.com and extmgmt.dev.azure.com only, as you, with read requests. It never changes anything in Azure DevOps, and nothing is written to your disk.
Work item titles, descriptions, comments and tags, and who an item is assigned to. Licence settings are never read either.
Counts, dates and work item ids only, never titles or names.
No licence check and no telemetry: the connector sends nothing to Baytek. What your assistant does with the numbers is between you and your assistant’s provider.
The admin switch
Configuration → General, “Allow the local AI connector”. It is on by default, and only Agile Analytics admins can change it.
An organisation admin can refuse every local AI connector connection in Configuration → General. The connector checks this switch before every answer and replies “disabled by admin” when it is off. The switch does not remove anyone’s Azure DevOps access, which already reaches the same data.
Each change is recorded in the Admin activity log.
Troubleshooting
When the connector cannot answer, it replies with a short code instead of a number. Here is what each one means.
| Code | What it means |
|---|---|
| workflow-not-mapped | This project has no saved workflow mapping. An Agile Analytics admin opens Configuration → Workflow Mapping, checks it and saves it. |
| settings-unreadable | Your Agile Analytics settings could not be read. Check that Agile Analytics is installed in this organisation and that your token has Extension Data (Read), then ask again. The connector never treats an unreadable switch as allowed. |
| azure-cli-sign-in-required | Azure DevOps did not accept your sign-in, or there is none. Check that AZURE_DEVOPS_PAT is set, has not expired and has the three scopes, and restart your assistant; on the Azure CLI route, run az login again. |
| disabled-by-admin | An Agile Analytics admin has turned the connector off for your organisation. Ask them about the switch in Configuration → General. |
| ado-request-failed | An Azure DevOps request failed or timed out, for example while Azure DevOps was busy. Try again in a minute. |
| team-not-found | The project has no team by that name or id. Check the spelling of the organisation, project and team. |
| no-finished-sprints | The team has no finished sprint yet. Most answers use finished sprints only, never the running one. |
FAQ
No. It is part of your Agile Analytics subscription. There is no licence check: it answers only when the extension is installed and a workflow mapping is saved.
No. The connector runs on your own computer, reads only counts, dates and work item ids with your own Azure DevOps access, and sends nothing to Baytek.
Yes. An organisation admin can turn it off for everyone in Configuration → General with the “Allow the local AI connector” switch, and each change is recorded in the Admin activity log.
Yes. The connector uses the same metric code as the views and your saved workflow mapping, and where it cannot read what a view reads it leaves the number out rather than estimate it.
Need help?
Our support team is available Monday through Friday, 9 AM to 5 PM Pacific Time.