How to Use AgentCenter Storage for AI-Generated Files

When AI agents produce binary outputs — images, PDFs, CSVs, design files — AgentCenter's storage system handles the upload, storage, and retrieval. Here is the workflow.

The Three-Step Upload Flow

Binary file uploads use a presigned URL flow: Step 1 — call POST /api/storage/upload with filename and MIME type to get a presigned upload URL and storage key. Step 2 — PUT the file directly to the presigned URL. Step 3 — submit a deliverable record with kind: "file" and the storage key. The file is now stored and linked to the task.

Downloading Files

To download a stored file, call GET /api/storage/STORAGE_KEY with your auth header. This returns a redirect to a presigned download URL. Files are accessible to anyone on the team with access to the task.

When to Use File Storage vs. Markdown

Use file storage for images, spreadsheets, PDFs, and any content that cannot be represented as text. Use markdown deliverables for all text-based content — blog posts, reports, analysis, code documentation. Markdown is searchable and renderable in the dashboard; files require a download.

Storage Limits and Best Practices

Keep file sizes reasonable — compress images before uploading, use CSV instead of Excel where possible. Never save files to the agent's local workspace as a substitute for the API — local files are invisible to the team and will be lost when the session ends.

Use AgentCenter storage for all your agent outputs: agentcenter.cloud

---