File Attachments
Documentation and guides for OutcomeDev.
Upload files directly into your task conversations to give the AI agent visual context, reference documents, or code samples.
Overview
OutcomeDev supports multi-file attachments in both the initial task form and follow-up chat messages. Attached files are stored securely on Vercel Blob and made available to the AI agent as part of its context window.
How to Attach Files
There are three ways to attach files:
| Method | How |
|---|---|
| 📎 Attach button | Click the paperclip icon in the toolbar to open the file picker |
| Drag and drop | Drag files from your desktop directly onto the input area |
| Clipboard paste | Copy a screenshot or image and press Cmd+V / Ctrl+V in the input |
When files are selected, they appear as chips above the text input showing the filename, size, and a remove button (✕).
Supported File Types
| Category | Formats |
|---|---|
| Images | PNG, JPG/JPEG, WebP, GIF, SVG |
| Documents | PDF, TXT, Markdown, CSV, JSON, XML, YAML |
| Code | TypeScript, JavaScript, Python, Ruby, Go, Rust, Java, HTML, CSS, Shell |
Limits
- Max file size: 25 MB per file
- Max files per message: 10 files
- Retention: Files persist indefinitely (tied to your account)
How the Agent Uses Attachments
When you attach files to a message, the agent receives them as context:
- Images → Injected as visual context, allowing the agent to "see" screenshots, mockups, designs, error messages, or UI references.
- Documents & Code → Referenced as downloadable links with metadata (filename, type, size), which the agent can fetch and analyze.
- Sandbox-Native Access → All attached files are automatically downloaded into the sandbox's current directory (
./). This allows the agent to use local tools (likels,grep, orcat) to analyze your files directly without needing to download them first.
This is especially useful for:
- Sharing screenshots of bugs, designs, or UI that needs to be replicated
- Providing reference code from another project or documentation
- Attaching data files (CSV, JSON) for the agent to process or integrate
- Sending PDF specs or design documents for the agent to follow
Where Attachments Work
| Location | Supported |
|---|---|
| Task Form (new task) | ✅ |
| Follow-up Chat (existing task) | ✅ |
| Scheduled Tasks | ❌ (text-only) |
Display in Chat
Attachments appear inline in your message bubbles:
- Images render as clickable thumbnails — click to open full-size in a new tab
- Documents and code files render as compact file chips with the filename, size, and a download link
Tips
- You can attach files with or without a text message — sending files alone is supported
- Use
Cmd+Vto paste screenshots directly from your clipboard — no need to save the file first - The drag-and-drop zone highlights with a visual indicator when you drag files over the input area
- Files that exceed the 25 MB limit are rejected with a toast notification before upload