How Edits Work

When a write tool succeeds, here’s what happens:

• The diagram changes immediately. The resource, field, or connection is updated in the workspace, exactly as if you’d made the change on the canvas.
• Nothing deploys on its own. The change sits in the workspace like any unapplied edit. You (or a teammate) still need to run Plan and Apply in Paxper before it reaches your cloud account.
• Open editors update live. If someone has the workspace open:
  • In collaboration mode, the change appears on their canvas instantly, the same as a live edit from another teammate.
  • Otherwise, their editor reloads the latest diagram — or, if they have unsaved edits of their own, warns them instead of silently overwriting their work.
• It’s recorded in the audit log, marked “via MCP”, with the real Paxper user who made it as the actor — not a generic service account. Admins reviewing the audit log can always see which changes came from an assistant and who authorized them.
• It’s recorded in the workspace’s history as “Edited via MCP”, alongside its normal edit history.
• The cost estimate recomputes to reflect the new diagram, the same as any other edit.

Example

You ask your assistant: “Add an S3 bucket for logs to the platform workspace, versioning on.”

1. The assistant calls get_resource_schema for the S3 bucket type to see what fields it can set.
2. It calls add_resource with the bucket’s name and versioning setting.
3. The bucket appears on the platform workspace canvas immediately — for anyone with it open, and the next time anyone else opens it.
4. The change shows up in the audit log as “via MCP” under your account, and in the workspace’s history as “Edited via MCP”.
5. The workspace’s cost estimate updates to include the new bucket.
6. Nothing is created in AWS yet — someone still needs to click Plan, review it, and Apply.