One sign-in, one approval, one record across the family

What the products share today, what is built and switching on, and what is planned.

Each of our products has its own site, team and price. They also share one contract: shared context, then an evidence-backed decision, approved execution, a measured outcome and reviewed learning. Here is what connects them, and the state of each.

Live

• One family list naming every product.
• A context packet, by content digest, on the pre-edit note, each Ask turn and each Send to the agent, attached to the task when there is one.
• Catalog entries in PromptCompute for proposals, sequences and discounts.

Built, switching on

• Approvals through PromptCompute for rollbacks, fixes and memory writes, for a team that connects. Signed receipts for lessons are not switched on yet.
• Sales as Code’s signed weekly accuracy receipts, for each paid seat with a graded deal. The pooled accuracy figure publishes once a group has five orgs and twenty graded deals, and none has yet.
• The coding agents that build Sales as Code read the shared brain. A release that forecasts worse is built to file a lesson for review, and it has not run yet.
• Sign in with PromptReports ID. Signing in as an owner links the team and connects PromptCompute.
• Credentials for each product, sent over a signed back-channel when an organization links.
• Publishing provenance records and corrections to AI Source of Truth.
• The CommitOps link: customer promises on the change page, agent cost and change risk back to CommitOps, and a broken promise opening a case.

Planned

• One task id carried by every product, and the claim statuses in the other products.
• One Connect page on PromptReports ID for every product.
• One autonomy setting per action, where deny always wins.
• ContextFile as an optional source: the loop runs without it, and connecting it adds the agent’s context history.