Waste Ledger
Cost by user, team, workflow, model, and integration; flags unusual spend and cost-reduction opportunities.
Open source
We ship the tools we use internally. Built for teams running Claude at organization scale.
For enterprise AI leaders
AI spend reporting for the person who has to answer for it — runs entirely on your machine.
Enterprise Claude deployments generate cost, usage, and risk that management has to answer for. Chief AI Officer turns that raw telemetry into three reports a leadership team can actually act on — and processes everything locally.
Python 3.11+
View on GitHubCost by user, team, workflow, model, and integration; flags unusual spend and cost-reduction opportunities.
Where adoption is gaining traction, which teams are deriving sustained use, and where further enablement is needed.
Reliability, security, and governance concerns — weak answer coverage, failed routing, stale data, and workflows needing review.
Runs locally. No cloud dependency, no telemetry.
For plugin developers
A plugin is written instructions, not code with a build step — nothing goes red when its behavior quietly changes. These close that gap.
Let the agent grade the agent. Point it at your plugin, your test cases and your quality rubric. It runs a live Claude session against the plugin, judges the output against your criteria, and reports pass/fail. Wire it to fire on every commit and quality-checking stops being a chore you keep meaning to get to.
View on GitHubStop fixes from landing in only one copy. When a pull request touches a watched file, it finds the sibling copies and posts a verdict for each one — likely-applies, different-on-purpose, conflicts, or unclear. Read-only by design: it reports, it never edits.
View on GitHubHand a Claude session to a teammate. Work started in Claude often has to be carried on by someone else — tomorrow, on another machine, in another person's account. This generates a handoff plugin fitted to your team's own storage, with credential redaction built in.
View on GitHub