The actuals of your
AI investment.
Aktuals turns every AI coding and agent session in your company into usage, productivity, and output intelligence — evidence of what AI work actually ships, for the engineers themselves, their leaders, and the people who sign the contracts.
One platform, three audiences
The same session-level evidence, surfaced differently for each person who needs it.
Your own picture
Your sessions, searchable. What AI handled, what you steered, where you're repeating yourself, what you could learn next. You always see everything collected about you — no exceptions.
Patterns, not rankings
Team-level work shape: output by category, effort per shipped artifact, efficiency and exploration mix, adoption depth. Cohorts and patterns by default — never leaderboards of individuals.
Renewal-ready evidence
Utilization against seats, effort trends, work linked to shipped artifacts — exportable evidence packs you can lay your own contract numbers over. No contract upload required.
How it works
AI coding tools already write a detailed record of every session to the machines they run on. Aktuals reads that record — locally — and turns it into intelligence.
A lightweight desktop agent
A single small binary on each machine reads AI session files and local git history. Processing happens on the device: parsing, categorization, secret redaction, privacy shaping — before anything leaves.
Derived data, by policy
Your organization chooses a privacy tier. The default ships metrics only — counts, durations, categories, confidence-banded links to shipped work. Raw prose stays on the machine unless your org explicitly opts into more.
Dashboards & evidence
Effort and output by team, category, and time. Session patterns. Adoption depth. Ask questions in plain language and get answers with confidence labels — backed by session-level evidence.
Built to be trusted by the people it observes
Analytics about engineers only works when engineers are comfortable with it. That principle is enforced in the product's architecture, not just its marketing.
Patterns and cohorts by default
Leadership views show how work happens — not ranked individuals. There is no "hide this from engineers" toggle, and there never will be.
Engineers see everything
Every person can see exactly what is collected about them, at which privacy tier, and who can view which aggregates. Transparency is a feature, not a setting.
Secrets never leave the machine
The agent runs credential redaction over every field before upload, at every tier. What it caught is auditable on the device.
Honest numbers
Estimates are labeled as estimates. Confidence is shown in bands, not false-precision percentages. Links between AI sessions and shipped work state their evidence.
Pricing
Aktuals is in its design-partner phase: a small group of teams shaping the product with us, on partner terms.
Design-partner program — now open
For teams of 10–100 using Claude Code and other AI agent tooling. Direct line to the founders, partner pricing locked in at GA.
About
Aktuals is built by the team behind AI Agent Usage, the on-device Mac app for AI cost and session insights. The same engine — parsing, attribution, insight analytics proven on real workloads — now built for teams, with privacy architecture designed for the enterprise. Actuals, in the finance sense: what actually happened, versus what was planned.
Get early access
Join the waitlist and we'll reach out as design-partner slots open.