Agents can read files you upload, create files for you to download, and work with files stored in their workspace. The Files page gives you direct access to My Files and every agent workspace you own.
Uploading files
Drag and drop a file into the chat, or use the attachment button. The agent can immediately read and work with it. Chat uploads go into My Files.
You can also upload individual files or folders from the Files page. Uploads are limited to My Files. To put an uploaded file in an agent workspace, copy or move it there after the upload finishes.
Common workflows:
- "Here's a CSV. Analyze it and give me a summary"
- "Read this PDF and extract the key points"
- "Take this JSON file and convert it to CSV"
Managing files in the Files page
The file tree has two roots:
- My Files contains your uploads and other user-owned files.
- Workspaces contains one folder for each agent you own.
You can create folders, rename items, copy, move, and delete inside either root. Copy and move also work between My Files and an agent workspace, or between two agent workspaces.
The roots themselves cannot be renamed, copied, moved, or deleted. Frona also checks ownership and rejects a path that points to another user's files or an agent you do not own.
Copy and move handle existing destination names differently:
- Copy keeps the existing file and gives the new copy a numbered name such as
report-1.pdf. - Move reports the conflict without moving any item in that request.
The current folder is stored in the page URL. Reloading the page or using the browser's back and forward buttons restores that folder, including nested agent workspace folders.
Getting files from agents
When an agent creates a file such as a report, script, or processed dataset, it appears as a downloadable attachment in the chat. The same file remains available in that agent's folder under Files → Workspaces.
Ask for files explicitly:
- "Write a Python script and save it as a file I can download"
- "Create a CSV report from this data"
- "Generate a summary and save it as a markdown file"
How agents access files
Agents use a small set of typed file tools for everyday file work:
| Tool | What it does |
|---|---|
read | Read a file (or a slice of one) or list a directory. |
write | Create or overwrite a file atomically. |
edit | Apply line-based inserts, deletes, or replacements to an existing file. |
glob | Find files matching a pattern. |
grep | Search file contents with a regex. |
These tools start in the agent's own workspace. A sandbox policy can grant access to user files with user://<user_handle>/<path> or to another owned agent workspace with agent://<agent_handle>/<path>. Frona resolves both forms inside the authenticated user's storage tree before it applies the sandbox.
Virtual paths are policy configuration, not file-tool input. Do not pass a raw user:// or agent:// URI to read, write, edit, glob, or grep. For the full path grammar and a Cedar example, see Virtual paths.
The agent can also use shell, Python, or Node.js when it needs a library or a multi-step processing pipeline.
To share a file with you, agents use the produce_file tool, which makes the file available as a downloadable attachment in the chat. Files live in the agent's workspace directory and persist across conversations.
Tips
- Name your files clearly. Agents use filenames to understand what a file contains. "q2-sales-data.csv" is better than "data.csv."
- Use files for large content. Instead of pasting a huge block of text into chat, upload it as a file. Agents handle files more reliably than very long messages.
- Files persist. Files in My Files and agent workspaces survive across conversations. You can reference them later: "Use the report from last time."
- Ask for specific formats. "Save it as CSV" or "Create a PDF" when you need a particular output format.
Next steps
- Agent Workspaces. How workspace file storage works.
- Running Python Code. Process files with Python scripts.
- Running Shell Commands. Use command-line tools on files.