Reference
Know where your data goes.
Understand API caching, action records, optional change history, custom content and temporary media storage.
On this page
How your data flows
You ask a question in your chosen AI client. The client calls Ad Superpowers with your authenticated connection. Ad Superpowers checks the workspace and account access, uses the relevant platform connection and returns the tool result to that client. Your AI client processes the result to answer you.
Ad Superpowers operates the connection and tools; it does not operate an AI model to generate your analysis. The client and model provider you choose have their own retention, training and access policies. Results, profile context or custom instructions returned by a tool can enter that client's conversation.
What Ad Superpowers stores
Different functions need different stores. The short API-cache window is not a retention promise for all data.
| Data | Why it is stored | Relevant boundary |
|---|---|---|
| Account, workspace, membership and connection records | Sign-in, access and operation of the service | Managed through the account/workspace lifecycle |
| Platform credentials | Make authorized requests to connected platforms | Encrypted in the application; access can be renewed or disconnected |
| API response cache | Reduce repeated platform requests | Short-lived expiry described below |
| Usage and internal action records | Usage accounting, investigation and attribution | Separate retention from the cache |
| Optional external change history | Keep supported platform change events for reporting | 365 days from event time |
| Client profiles, custom skills and workflows | Keep the business context and reusable content you save | Stored content, not temporary response caching |
| Uploaded media and upload-result evidence | Transfer a file to the selected platform and track the outcome | Separate source-file and result windows |
API response caching
Retrieved API responses can be cached for one minute to six hours, depending on the endpoint and data period. Account lists have their own cache; the current MCP assignment-cache limitation can delay changes in account access by up to 30 minutes. See team permissions.
Cache expiry does not remove copies already received by your AI client. It also does not delete a saved client profile, custom template or event in the optional changes archive.
Usage and internal action records
Usage records identify tool usage for accounting and investigation. The workspace activity trail records supported Ad Superpowers actions with the authenticated actor, time, target, outcome and summary. It records changed-field names for supported operations rather than serving as a complete before/after copy of campaign data.
Usage logs, workflow run logs, email logs, the workspace activity trail and administrative and security audit records (including sign-in events) have a default retention of 365 days. Session records are deleted 30 days after the session ends. Metering records, which count tool calls against a workspace's plan, are kept for as long as the workspace exists. When you delete your account, workflow run logs and session records are deleted with it; the other records are kept without a link to your account until their retention ends. Applicable legal holds can affect deletion of internal records.
Workflow run records store a parameter count rather than the supplied parameter names and values. This does not mean the workflow template or other content you explicitly save is transient.
When teammates use a shared platform connection, Ad Superpowers can attribute its recorded action to the authenticated teammate. A platform's own event can identify the platform connection's authorizing user instead. Do not automatically equate those identities.
Optional external change history
Logging archives supported Google Ads, Meta and selected GA4 configuration events after a source is enabled and import succeeds. A workspace default can enable Google Ads and Meta for new dashboard clients; GA4 requires a separate client choice and additional permission. See change history for controls and coverage limits.
The archive can retain account/object identifiers, event times, operations, source information and coverage. Allowed configuration details can include advertising text, targeting, keywords, budgets, bids and status. Google/GA4 actor addresses, Meta actor identity and permitted old/new details are encrypted in the application.
Credentials, passwords, tokens, individual lead/conversion data, Customer Match contents and unapproved detail fields are excluded or masked. URL query strings and fragments are stripped from permitted URL details. A missing detail is not evidence that the value never changed.
Owners/admins can read and export this archive. Managing collection requires an owner/admin with write permission. External events are retained for 365 days from their event time; pausing does not extend that period. Removing a client or platform connection does not delete the account archive. Workspace deletion removes the workspace's archive. A client filter uses current profile account links.
Client profiles and custom content
Client profiles store the goals, budgets, linked account IDs and other context your workspace enters. Profiles are workspace content, accessible to its members; they are not per-person private notes or a substitute for account permissions.
Custom skills and workflows store the content you create, including any client information you put in their instructions. Their visibility and publication settings determine who can discover and use them. Organization content belongs to its workspace; personal content belongs to its creator. Publicly sharing a skill can disclose the text to other users.
Review content before saving or sharing it. Keep credentials out of profile notes, skills and workflows. Disabling or archiving an item changes its availability and can leave its stored content in place. Use the available content-management and deletion controls, or contact support for a deletion request. See skills and workflows.
Temporary private media uploads
When your workspace has upload access to the selected destination, your client sends original media through Ad Superpowers to temporary private Supabase Storage and then to the selected platform. The upload route supports JPEG/PNG images and MP4 video for supported destinations; check media upload for current platform, file and size requirements.
Embedded metadata travels with the file; the transport does not automatically remove it. Only the HTTP byte transfer stays outside MCP payloads. Filenames and results can reach your AI client, as can files you already gave it or legacy base64 inputs. File access depends on your client; a remote MCP server cannot open a local attachment by itself.
Upload URLs expire after ten minutes. Finalization is available for 24 hours from reservation. Source objects are scheduled for deletion after completion, failure or expiry. The 15-minute cleanup cadence and daily bounded orphan scan are not guarantees of physical deletion at an exact deadline.
Stage/result evidence has a separate 30-day window and is pruned after source deletion. Audit records, database backups, local originals and platform copies are separate. Database backups do not include Storage media bytes. This temporary route is not a permanent media library or backup; keep your own originals.
Your AI provider's policies
The terms for your chosen client, model, subscription and settings govern its handling of conversations and tool results. Ad Superpowers' cache expiry does not set the AI provider's retention or training policy.
Check the provider's current documentation: Anthropic privacy, OpenAI data controls or the Gemini privacy hub. For clients that let you choose a model provider, check both the client and the provider you configured.
Choose the information you send
Keep requests specific: the relevant account, period, fields and task. Connecting through MCP supplies an authenticated access path; it does not make all downstream processing private by default. Copying a report into another tool, exporting a CSV or publicly sharing a skill creates another disclosure to review.
Use your own identity, restrict account access where appropriate and set read-only permission for people who only need reports. See workspace permissions.
Security controls
Sensitive fields such as platform OAuth tokens and client-profile content use authenticated application encryption with Fernet. Other records use different protection layers; this is not a claim that every database field is separately encrypted by the application.
Workspace access checks, role checks and write permissions govern tool use. Protect your login, API keys and upload URLs, and revoke or rotate credentials when appropriate. The security page describes the wider controls and their scope.
For agencies managing client data
Agree which client data may be processed, which AI client and plan your team uses, and who may access each workspace. Keep client profiles and shared templates within that agreed scope. Check how the AI provider handles tool results before connecting confidential accounts.
The privacy policy, terms and sub-processor list describe the published policies and service providers. This guide explains product behavior; it does not replace those documents or your client agreements. Contact support for the applicable data-processing paperwork or a specific deletion request.
Learn more
- Change history for enabling, pausing, viewing and exporting external events.
- Media upload for transfer and retry behavior.
- Accounts and workspaces for access and context.
- Contact support for a question about stored data.
