Storage

Why storage matters

Apps SDK handles conversation state automatically, but most real-world apps also need durable storage. You might cache fetched data, keep track of user preferences, or persist artifacts created inside a component. Choosing the right storage model upfront keeps your connector fast, reliable, and compliant.

Bring your own backend

If you already run an API or need multi-user collaboration, integrate with your existing storage layer. In this model:

• Authenticate the user via OAuth (see Authentication) so you can map ChatGPT identities to your internal accounts.
• Use your backend’s APIs to fetch and mutate data. Keep latency low; users expect components to render in a few hundred milliseconds.
• Return sufficient structured content so the model can understand the data even if the component fails to load.

When you roll your own storage, plan for:

• Data residency and compliance – ensure you have agreements in place before transferring PII or regulated data.
• Rate limits – protect your APIs against bursty traffic from model retries or multiple active components.
• Versioning – include schema versions in stored objects so you can migrate them without breaking existing conversations.

Persisting component state

Regardless of where you store authoritative data, design a clear state contract:

• Use window.openai.setWidgetState for ephemeral UI state (selected tab, collapsed sections). This state travels with the conversation and is ideal for restoring context after a follow-up prompt.
• Persist durable artifacts in your backend or the managed storage layer. Include identifiers in both the structured content and the widgetState payload so you can correlate them later.
• Handle merge conflicts gracefully: if another user updates the underlying data, refresh the component via a follow-up tool call and explain the change in the chat transcript.

Operational tips

• Backups and monitoring – treat MCP traffic like any other API. Log tool calls with correlation IDs and monitor for error spikes.
• Data retention – set clear policies for how long you keep user data and how users can revoke access.
• Dogfood first – run the storage path with internal testers before launching broadly so you can validate quotas, schema evolutions, and replay scenarios.

With a storage strategy in place you can safely handle read and write scenarios without compromising user trust.