Release notes

Every update to Milestone, newest first. What changed, and who it is for.

Last updated

Showing all 59 updates

October 2026

11 updates

NewPreview

Playbooks: manager performance deep dive and leadership AI impact briefing

Two new one-click analyses join Playbooks. The manager deep dive compares a team’s developers with a wider baseline and highlights who is worth a closer look. The AI impact briefing summarizes AI’s effect for leadership, with a board version that names no individuals or teams. Every Playbook report shares one layout: status, what to do, key figures and details, with PDF and text export.

For Engineering managers and executives preparing performance reviews or board updates, who want a consistent readout without assembling it from dashboards.

New

Data Collection Agent changelog and update notices

An Agent changelog in Settings > Integrations shows what changed in each release of the Data Collection Agent, filterable by integration. When a self-hosted agent falls behind, admins see a notice listing only the changes relevant to their integrations, with per-host versions and upgrade guides.

For Admins who host the Data Collection Agent themselves and want to know when to upgrade and why.

New

AI Prints installers for device management

AI Prints now ships macOS .pkg and Windows MSI installers ready for MDM rollout, with published SHA-256 checksums for every download and a software bill of materials for each release. One-line install scripts remain available for individual developers.

For IT and platform teams rolling AI Prints out across an engineering organization and through security review.

Improved

See how the assistant read every data answer

Every data answer from the AI Assistant ends with a consistent summary of the time window, scope, population, metric definition, data coverage and permissions behind it. Follow-up answers shorten it to one line naming only what changed.

For Anyone who shares assistant numbers with others and needs to know exactly what a figure includes.

New

Agents discovered automatically, with numbers that follow your scope

Coding and review agents now join your Agents fleet automatically, marked unreviewed until an admin reviews them. Agent pages and summary cards follow your team, product or repository scope. Gemini Code Assist and Google Jules are now recognized as agents.

For Leaders tracking a fleet of AI agents, who want new agents to appear without setup and scoped numbers that count only their own agents.

Improved

GitHub Copilot spend by model

Copilot spend is now broken down by model, and usage covered by your seats’ included allowance appears as covered spend instead of disappearing.

For Finance and platform owners comparing AI tools, who need to see which Copilot models drive cost and how much value seats already deliver.

New

Pin assistant tables to Vibe Metrics

Tables the AI Assistant builds in chat can be pinned to a Vibe Metrics dashboard as live tiles, which refresh and respect each viewer’s own data access.

For Leaders who build dashboards by asking questions and want a good answer to become a shared, always-current tile.

NewPreview

Playbooks, starting with a month-to-date AI spend review

Playbooks introduces one-click analyses. The first is a month-to-date AI spend review: run it and get a report with clear verdicts and a permanent link. Where data is missing, the report says so instead of showing zero.

For Budget owners who need a monthly AI spend readout they can forward with confidence.

Improved

Agents page redesign

The Agents page brings every agent into one table, including idle ones, with last activity, repositories, tokens and lines changed. Each agent gets its own page with token usage, activity and review activity, and agents can be assigned to a product.

For Engineering leaders running coding and review agents, who want to spot idle agents and drill into any agent’s cost and output.

New

Web installer for self-hosted collectors

Customers running the Data Collection Agent on their own infrastructure get a browser-based installer: guided configuration with live validation, a review of changes before applying, preflight checks, Docker or Kubernetes deployment, and an Operate tab with status and a downloadable support bundle.

For IT teams at self-hosted customers who want to set up and troubleshoot collection without editing configuration files by hand.

Improved

The assistant shows how it read your question

AI Assistant answers now explain, in plain words, the metric, time window, filters and scope they used, with relative dates such as “last 30 days” resolved to real dates.

For Anyone who wants to catch a misread question before quoting the number.

September 2026

19 updates

New

Budget alerts in Slack

Milestone now sends AI budget alerts to Slack, delivered to the people responsible rather than a channel-wide ping. Admins can switch alerts off at any time.

For Anyone who owns an AI spend budget and would rather hear about overspend than discover it on a dashboard.

New

ChatGPT Enterprise support

ChatGPT Enterprise joins the supported AI tools, with per-user daily usage collected through an OpenAI admin key.

For Organizations whose AI use extends beyond coding tools and who want a complete picture of AI usage.

Improved

Ask the assistant about the page you’re on

The AI Assistant now knows which dashboard page you are looking at and shows it above the conversation, with chips to focus on a single widget. Questions like “why is this number down?” just work.

For Anyone who has had to describe what is on their screen before asking a question.

New

Connect self-hosted Jira

The Jira integration adds a Self Hosted mode for Jira Server and Data Center, configured directly from the Integrations page.

For Enterprises on Jira Server or Data Center who want to connect without help from Milestone.

Improved

Session duration in Session Explorer

Session Explorer adds a Duration column and Start date and Duration filters. Each session’s tokens, cost and duration now cover the whole session, so every figure on a row lines up.

For Anyone comparing AI coding sessions by length and cost.

New

BU Manager role

A new BU Manager role covers managers whose responsibilities span several teams and groups, scoping their view to exactly the units they manage.

For Large organizations whose management structure doesn’t fit a single hierarchy level.

New

Connect self-hosted GitLab and Bitbucket

GitLab Enterprise and Bitbucket Server and Data Center can now be connected from the Integrations page with your own base URL.

For Enterprises running self-hosted Git platforms who want self-service setup.

New

LiteLLM gateway spend

Milestone now reads AI spend directly from a LiteLLM gateway, without double counting cost already reported by individual tools.

For Organizations that route model traffic through LiteLLM, where the real AI bill lives.

Improved

Linear self-serve setup

Linear can now be connected from the Integrations page with just an API key. Issue-to-PR links and full issue status history come with it.

For Teams on Linear who want to connect it themselves.

Improved

Redesigned navigation

A new left navigation rail with grouped sections, flyout menus when collapsed and a rebuilt account menu.

For Every Milestone user.

New

Linear support

Linear joins Jira and Asana as a supported issue tracker, so cycle time, issue lifecycle and assistant questions work for Linear teams.

For Engineering organizations that plan their work in Linear.

Improved

Ask the assistant about your own work

The AI Assistant now knows who is asking, so questions like “how much have I spent?” or “what did I ship?” answer about you. It also answers questions about a pull request’s target branch and works the same across Jira, Asana and Linear.

For Every assistant user, especially anyone checking their own work.

New

Insights: pull requests merged without human review

A new Insights detector flags pull requests merged without a human reviewer, telling human reviewers apart from AI agents, by team.

For Engineering leaders who want to know whether AI-written code is landing unreviewed.

Improved

GitHub Copilot Chat in Session Intelligence

Copilot Chat now feeds Session Intelligence, Explore and the cost views on the same footing as every other AI tool.

For Organizations standardized on GitHub Copilot.

Improved

Scope a dashboard to several teams at once

Select up to 20 teams, groups, sites, repositories or products at once and see data for everyone in them. Saved views and bookmarks keep working.

For Leaders who own more than one team and have been adding up numbers by hand.

Improved

AI Spend Hub gets its own section

AI Spend Hub becomes its own navigation section with Overview, AI Efficiency and Budgets, and tool selection becomes multi-select, so you can see combined spend across the tools you choose.

For Finance and platform leads managing several AI tools at once.

Improved

Per-PR AI cost from any telemetry source

Per-pull-request AI cost now appears whenever cost data exists from any source, no longer only with AI Prints installed.

For Customers using session telemetry who want per-PR cost without extra setup.

Improved

Cursor sessions mapped to developers

Admins can add a Cursor Organization Admin key so Cursor session, skill and tool activity is attributed to the right developers.

For Organizations using Cursor who want every session tied to a person.

Improved

Contributor edits apply immediately

Excluding, reclassifying or confirming a contributor now shows up on your next page refresh.

For Admins keeping contributor data clean.

August 2026

8 updates

New

Session Intelligence

A new page shows what happens inside AI coding sessions. Explore slices activity session by session, Session Profile shows outcomes, spend, skills and MCP calls per developer and team, and a Skills Leaderboard shows which skills are used most.

For Engineering leaders who see aggregate AI spend but want to know who uses which skills, and at what cost.

Improved

Install the Milestone Slack app yourself

Admins can now install the Milestone Slack app from the Integrations page, with a guided setup and clear install steps.

For Admins who want Milestone in Slack without waiting on our team.

New

Cursor telemetry

Cursor joins the AI telemetry integrations with self-serve setup, so Cursor adoption, sessions and skills read the same way as Claude Code, Copilot, OpenCode and Codex.

For Organizations standardized on Cursor.

Improved

Drill into Vibe Metrics charts

Click any point on a Vibe Metrics chart to see the rows behind it, filtered to that time period and dimension.

For Anyone who has spotted a spike and wanted to know which PRs caused it.

Improved

AI Efficiency: total spend, costliest PR and a cost-per-PR benchmark

AI Efficiency now leads with total AI spend against the prior period and your costliest pull request, plus a median spend-per-PR benchmark for comparing developers.

For Engineering leaders and finance partners preparing budget conversations.

New

Ask Milestone from Slack

Milestone in Slack answers with native Slack charts, a single live progress line and downloadable PDF reports.

For Teams who live in Slack and want dashboard-quality answers without opening Milestone.

New

AI cost per merged PR

A new Merged PRs table ranks merged pull requests by the AI cost attributed to them, with per-model detail and an organization median. Budget & Usage now lists every developer and is searchable.

For Owners of the AI tools budget who want to know what a given PR cost.

New

OpenCode and Codex attribution

OpenCode joins the tools Milestone attributes commits and adoption to, and Codex attribution now reads exact commits and pull requests from Codex telemetry.

For Teams running a mixed AI toolchain.

July 2026

14 updates

New

Agent spend and review impact

The Agents page shows spend for review agents, including seat-based ones, and a new Comments acted upon metric: the share of an agent’s review comments that led to code changes.

For Teams evaluating coding and review agents on cost and real impact.

Improved

New global filters

A global scope bar with an entity picker and a timeframe control gives one consistent place to set scope and dates across the product.

For Every user who filters by team and date.

New

Use your own data in Vibe Metrics and the assistant

Vibe Metrics and the AI Assistant can now work with your own custom tables alongside Milestone’s data.

For Customers who want their own data reflected in metrics and answers.

New

Self-serve DORA metrics

Generate an ingest token and push deployment events from your CI/CD using in-product recipes. Deployment frequency, lead time, change failure rate and MTTR then become available as metrics.

For Teams who want DORA metrics without an integration project.

New

AI Efficiency view

AI Spend Hub and developer profiles show average cost per PR against organization and scope medians, with a Tokens column beside spend.

For Engineering leaders judging whether AI spend is efficient.

Improved

AI telemetry connector health

Integration cards for Claude Code, Codex, Copilot and LiteLLM telemetry now show whether data is fully flowing, partly flowing or not connected, and list the users discovered in the telemetry.

For Admins setting up AI telemetry who want to confirm it works.

Improved

Smarter developer filters and custom attributes

Attribute filters now combine the way you’d expect (any value within an attribute, all attributes together), and admins can define their own contributor attributes.

For Admins segmenting developers by location, level or business unit.

New

Research reports from the assistant

The AI Assistant can produce longer research reports in chat, export them to PDF, and reference developers, teams and repositories as clickable chips.

For Anyone using the assistant to investigate a question and share the result.

Improved

PR and issue titles across the product

Pull request and Jira issue titles now appear on hover wherever a PR or issue is listed.

For Anyone reading tables of PRs or issues.

New

Claude Enterprise spend in AI Spend

Usage and cost from Claude Enterprise now flow into AI spend reporting.

For Customers using Claude Enterprise who want complete AI spend totals.

New

A chat-first homepage

The homepage now centers on the AI Assistant, with a personal greeting and suggested questions that open an answer in one click.

For Every user, especially anyone who just wants to ask a question.

New

Amazon Kiro support

Amazon Kiro adoption and spend are now tracked and attributed to developers, and Kiro and Augment are included in per-PR AI cost.

For Teams using or evaluating Amazon Kiro.

New

Cost efficiency trend

AI Spend Hub and developer profiles add a Cost Efficiency Trend showing average AI cost per PR over time.

For Engineering leaders tracking whether AI spend turns into delivery.

Improved

Ask the assistant about sessions, skills and tools

The AI Assistant can now answer questions about AI coding sessions: running time, cost, tokens, and which tools, MCP servers, skills and subagents were used.

For Anyone who wants to understand how AI tools are actually used, not just what they cost.

June 2026

7 updates

New

AI Spend Hub

A dedicated section for AI tool spend, with a breakdown view and per-PR delivery columns.

For Engineering leaders and finance who want the AI bill in one place, mapped to delivery.

New

Agents page and cost roster

A rebuilt Agents overview with fleet KPIs, category tables and a per-agent roster with total cost, all following your team filter. Agents can be registered manually and confirmed in bulk.

For Teams tracking coding agent adoption and cost.

Improved

AI Assistant upgrades

Build new metrics from a prompt with the AI Metric Builder, and pin important conversations to the top.

For Anyone using the assistant to explore metrics.

Improved

The assistant respects roles and permissions

The AI Assistant enforces the same role-based data access as the rest of Milestone, so every user sees only the data they’re allowed to.

For Admins at organizations with scoped roles.

Improved

Table view in Vibe Metrics

Switch any metric between its chart and a sortable, searchable table, in the editor and on dashboard cards.

For Anyone who needs exact numbers, not just a chart.

New

Claude Enterprise and org-wide Claude Code telemetry

Connect Claude Enterprise as its own integration, and roll out Claude Code telemetry to every developer at once from the Claude admin console.

For Admins onboarding AI tools across an organization.

Improved

More accurate per-PR AI cost

Per-PR AI cost now separates seat-covered usage from pay-per-token spend for Claude and Codex, so it reflects how the tools are actually billed.

For Anyone attributing AI cost to delivery.

About Milestone

Milestone is the intelligence platform for AI native engineering. Engineering leaders use it to measure AI spend, agents and engineering ROI across AI coding tools such as Claude Code, Cursor, GitHub Copilot and Codex.

Back to top