Do you store copies of my ad platform data?

Performance data is read on demand per task rather than maintained as a copy. What is stored is task history and output, account memory, dashboard snapshots and encrypted credentials — all scoped to one organization.

Are my emails stored when I use Agent Mail?

No. Message bodies are re-fetched on demand rather than stored, and a read is checked against your organization before anything is returned.

What is deleted when I disconnect an integration?

The stored credential rows are deleted and the cached resolved token is invalidated at the same time. For Shopify, an uninstall triggers a purge of every stored connection for that shop via Shopify's compliance webhook.

Can I keep task data away from third-party model providers?

Task content necessarily reaches a model to be processed. Bring Your Own Model routes those calls to an OpenAI-compatible endpoint you control instead of the built-in gateway.

Trust

Data handling: what we store, what we don't, and how deletion works

TL;DR

Most of what the agent works with is read on demand from your platforms rather than copied into a warehouse of ours. Where we do persist things — task history, account memory, dashboard snapshots — it is scoped to one organization, and deletion paths are real rather than a flag on a row.

Read on demand vs stored

Data	Handling
Ad platform performance data	Read on demand for each task; not maintained as a copy
Email message bodies (Agent Mail)	Not stored — re-fetched on demand and ownership-checked before being returned
OAuth credentials	Stored, encrypted, scoped to one organization
Task history and outputs	Stored so you can revisit a run
Account memory (targets, protected campaigns, rules)	Stored — the point is that it persists across runs
Dashboard snapshots	Stored, last ten kept per dashboard

Organization isolation

An organization is the isolation boundary, not a label. Connected accounts, integrations, account memory, skills, audit log and billing all belong to one organization, and a request for data outside the caller's organization is rejected rather than filtered. Agencies use this to keep each client structurally separate — see Agent Planners for agencies.

Deletion

Disconnecting an integration deletes its stored credentials and invalidates the cached token at the same time.
Shopify shop deletion is wired to Shopify's own compliance webhook: when a shop uninstalls, every stored connection for that shop is purged, and a failed purge is surfaced loudly rather than swallowed.
Data subject requests — a merchant asking what is held about a specific customer is answered against the fact that we do not key data by their customers.

See compliance for the regulatory framing, and the privacy policy for the formal statement.

Model providers

Task content passes to model providers through the AI Gateway in order to be processed — that is inherent to using a model. If that routing is unacceptable for your data, Bring Your Own Model points every agent's LLM calls at an endpoint you control instead.

Frequently asked questions

Do you store copies of my ad platform data?: Performance data is read on demand per task rather than maintained as a copy. What is stored is task history and output, account memory, dashboard snapshots and encrypted credentials — all scoped to one organization.
Are my emails stored when I use Agent Mail?: No. Message bodies are re-fetched on demand rather than stored, and a read is checked against your organization before anything is returned.
What is deleted when I disconnect an integration?: The stored credential rows are deleted and the cached resolved token is invalidated at the same time. For Shopify, an uninstall triggers a purge of every stored connection for that shop via Shopify's compliance webhook.
Can I keep task data away from third-party model providers?: Task content necessarily reaches a model to be processed. Bring Your Own Model routes those calls to an OpenAI-compatible endpoint you control instead of the built-in gateway.