What we won't build
Engineering analytics has an obvious way to go wrong: turn people into numbers and hand the numbers to their manager. These are the specific things we have decided not to do — and one place where the line is blurrier than we would like.
We do not score developers
There is no productivity score, no performance rating, and no composite number attached to a person. Nothing in Acticly computes one, and we are not going to add one. The only percentile ranking in the product compares your team against external DORA benchmarks — never one colleague against another.
A single number per person is the fastest way to make a measurement tool into a management weapon, and the number is always wrong anyway.
We do not read your code or your prompts
Acticly receives model names, token counts, and timing. Not prompts, not completions, not source code. The Claude Code and Codex setups we publish set prompt logging off and force the trace exporter off — you can read the exact configuration on the integration pages before you install anything.
Spend analysis does not require the content, so we do not take it. Anything we hold is something we could leak.
We do not write to your systems
Every code and issue integration is read-only. Acticly never pushes code, opens pull requests, creates branches, transitions Jira issues, or comments on anything. If you revoke our access tomorrow, nothing we did remains in your repositories.
A reporting tool with write access is a reporting tool that can break your build at 2am.
We do not act on your spend unless you turn it on
Spend controls ship in dry-run for every organization. The loop computes what it would do and writes it to an audit log, and nothing reaches your gateway until someone deliberately enables enforcement. There is no default that blocks a developer mid-task.
A tool that can cut off your engineers should have to be switched on by a person who understood the decision.
Where this gets uncomfortable
We do show AI cost per person, and today anyone in your workspace can see it. That is a real cross-person comparison, and we are not going to describe it as anything else.
The reason is operational rather than philosophical: a budget needs an owner and an alert needs a name. "Someone spent $400 yesterday" is not something anyone can act on. Cost lands on a person because that is the only way the alert is useful.
But cost per person and a performance ranking are different things, and the difference only holds if we keep it. So: no score is derived from it, it is never combined into a rating, and it is not used to compare people on anything except spend.
If you want that view restricted to admins, say so. It is a permission model, not a principle, and nobody has asked us for it yet.
What we do instead
Delivery reporting is aggregated at the team level. Individual names appear where an operational decision needs one — a spend alert, a budget owner, an identity to map — and nowhere else. If you find a screen that contradicts anything on this page, tell us and we will either fix the screen or fix the page.
See what the two products actually do · What Acticly reads from each tool