Memory apps

A memory app is a published page over your team's memory: a table, a pipeline, a scorecard, a timeline. The layout and rules are fixed when you publish. The data keeps updating as your team works.

The settings page and the legal terms call the feature Memory-Native Apps.

Where apps live

Apps live in your team's workspace, at /apps on your workspace address. Each app has its own address under /apps. Each app belongs to one environment, such as production or a sandbox, and reads only that environment's memory.

Apps are available only inside the workspace. There is no public link.

What an app is made of

An app is stored as a specification. Achiral does not generate or run code for it. The specification holds:

• the connected sources the app reads, called its bound sources;
• the entity types it is about;
• your signals, source authority rules, formulas, and stages;
• pages built from eight components: table, scorecard, timeline, pipeline, narrative, evidence drawer, filters, and review queue; and
• the actions it offers.

Achiral validates the specification before it is saved. Unknown components, properties, and actions are rejected.

Who can do what

Task	Owner or Admin	Developer or Member
Build an app and preview it	Yes	Yes
Edit an app someone else built	Yes	No
Submit an app for its first publication	Yes	Yes
Publish an app for the first time, or approve a submission	Yes	No
Publish a later release of an app they steward	Yes	Yes
Change who can open an app they steward	Yes	Yes
Correct a fact or record a decision through an app	Yes	No
Turn the add-on on or off	Yes	No

The member who builds an app is its steward. The team owns the app. An Owner or Admin can reassign the steward.

Chapters

• Building an app: how to start a memory app from a description or with the structured builder, preview it, and name it.
• Signals, formulas, and stages: how to define what a memory app reads and computes, and confirm each definition before publishing.
• Run limits: the fixed limits on each run of a memory app.
• Publishing an app: who can publish a memory app, how first-publication approval works, and why publishing can fail.
• Audiences and stewards: who can open a published memory app, how invitations work, and who stewards it.
• Versions and rollback: how revisions, releases, rollback, and archiving work for a memory app.
• Pricing and the add-on: what the memory apps add-on costs, what is billed, and what happens when you cancel.
• Evidence and value status: how each value in a memory app is marked, and how to open the evidence behind it.
• Confidence and the review queue: how the confidence setting decides which inferences count, and how the review queue works.
• Component states and narratives: what each component state means, what happens when a source disconnects, and how narratives are generated.