AI Context and Maintenance

AI Context

Create one stable, versioned project URL for trusted AI chats, IDE tools, and team workflows.

AI Context is a read-only JSON snapshot of the project’s semantic planning. It replaces manual whole-project JSON downloads and Markdown context packages for ongoing AI work.

  1. Open a project and wait for Saved.
  2. Choose AI Context → Manage AI Context.
  3. Choose Create AI Context link.
  4. Copy the URL into the trusted AI chat or IDE that needs the plan.

Creation publishes revision 1. The URL belongs to the project and is shared across authorised editors; another user does not generate a different link.

What the URL contains

The snapshot organises planning into the following categories. These are compatibility labels in the AI format, not separate workspaces in the current editor:

  • Project Overview
  • Structure
  • User Journeys
  • Sections
  • Page Hierarchy
  • Technical Architecture
  • Design Guidance
  • Pre-flight

Project Resources and Design References are represented by their titles, descriptions where available, and public HTTP(S) URLs. Image and document binaries are not embedded. AI tools can follow those links when they have web access; an inaccessible or private link should be reported rather than treated as read.

Documentation supplies written project and design context; Structure combines the graph categories; Pages supplies reusable definitions and page composition. The snapshot is a planning representation, not a complete backup of the live editor. Its omission of content is not an instruction to delete that content.

It also identifies its context revision and the version of its read-only AI workflow contract.

AI Context includes bounded frame-local node bounds and semantic connector handles as planning reference. An editing agent still needs to inspect the current canvas before making changes. It excludes raw canvas state, database records, collaboration presence, locks, edge bend points, styles, and internal persistence state.

Publish project changes

Editing the project does not silently change the public snapshot. This lets the team finish and review a set of changes before publishing them to AI tools.

When saved project changes are newer than AI Context, the AI Context menu shows an unpublished-changes indicator and offers Update AI Context. You can also publish through Manage AI Context. Choose it to publish the current semantic plan.

The URL stays the same. A meaningful context change creates a new revision; changes that do not affect semantic AI context simply mark the existing snapshot as checked and current.

Use it with an AI tool

Provide the URL as planning reference. It contains the most recently published snapshot, which may differ from the current editor.

Read this AI Context for background, then use computer use to inspect the
current Structure workspace and update the confirmed CMS integration.
Preserve unrelated nodes and connections and verify the change saved.

The agent needs browser controls and access to the signed-in editor to make changes. Without them, it can suggest precise manual edits. Existing-project bulk update files are no longer supported; substantial rebuilds use a seed to create a separate project.

Security

AI Context is a secret bearer link:

  • Anyone with the URL can read the published project context.
  • Share it only with trusted people and tools.
  • Do not place it in public tickets, documentation, source code, or screenshots.
  • Skafold rate limits public access and serves cacheable immutable snapshot content to reduce repeated origin work.

Deleting the link revokes the existing URL and can break team members’ AI chats or IDE configuration. Skafold requires a second confirmation and reminds you to inform the team. After deletion, creating AI Context produces a different URL.

AI Context is separate from the client Share link. Client sharing provides a view-only visual project; AI Context provides structured machine-readable planning.

Continue with Keep AI Context current.