Pilots

A pilot is a team running the checks on its own records, on its own machines, and signing the report it keeps. Nothing of the team's leaves its machines unless the team sends it.

What you get

• Record, for any single Claude Code log, behind the Desk password until it is fully vetted.
• The census tools, to run over every log a team holds: counts by month and by project, own sessions apart from subagents, every write outside the folder listed; the report signed with the team's own key; verified offline by anyone with the public key; anchored in a public log if the team chooses. The tools are not public; ask us.
• The published receipts and their kits, for the quantitative questions, with the one command each.
• A rule written down for each check, with Repeat's and Scope's limits printed on every receipt. A rule that cannot run on your format is taken out, not left in.

What you bring

• Your logs, kept where they are. The census reads a folder; it uploads nothing.
• One person who holds the team's signing key.
• The question you want answered, in your words, so the rule can be written down before the run. If your agents write a format we have no rule for, writing that rule is the first piece of work, and the receipt says so.

What you keep

Every receipt and report, signed by your key. The public key to verify them. The commands that reproduce every number. If you anchor a report, the public log entry is yours to point at.

What is not in a pilot

No access to your systems. No language model run over your logs. No verdict on your agents or your people. No promise about a format we have not measured.

Start

Write to hello@thecommons.ooo with which agent you run and how many logs a month. Or start without us: check one log now, then run the census.