Agent files: its "mind" and its outputs Guide

Every Praxivara agent keeps files, and they come in two flavors: the files that make up the agent's own "mind," and the files it produces as it works. You'll find both on the agent's Files tab, under the Configure section of its detail page.

The two kinds of files

The Files tab separates what an agent is from what an agent makes. Switch between the two groups to see each:

• Agent files — the agent's persona and memory. These describe who the agent is, what its purpose is, what it knows about you, and what it has learned over time.
• Output files — everything the agent has produced while doing its job: reports, PDFs, images, and other deliverables.

Agent files: the agent's "mind"

These files are how an agent stays consistent from one run to the next. Rather than starting from scratch every time, the agent carries forward a sense of its identity, its purpose, and what it has picked up along the way.

Typically you'll see a few kinds:

• Identity and purpose — who the agent is and the job it was set up to do.
• Memory — details it has learned about you, your business, and your preferences, plus useful facts it has gathered while working.

Reading these files is a great way to understand how your agent thinks and why it makes the choices it does. They're there for you to open, and reading them changes nothing about the agent's behavior.

Note: Agent files are created on the agent's first run. If you've just built an agent and it hasn't run yet, this section may be empty. Run a quick test or let its first task complete, then check back.

Output files: what the agent produces

Whenever an agent generates something — a formatted report, a PDF, a chart, an image — that file lands in the Output files group. You can browse everything the agent has ever produced in one place.

Output files also appear on the agent's Deliveries tab, next to the run that produced them. The Files tab simply gathers them together so you can find and download older work without hunting through individual runs.

Opening, previewing, and downloading

Both kinds of files work the same way:

1. Open the agent's Files tab.
2. Choose Agent files or Output files.
3. Select a file to preview it in place.
4. Use the download option to save a copy to your device.

Files are read-only for you

You can open, preview, and download any file, but you can't edit files directly from this tab. Keeping them read-only protects the agent's memory and its finished work from accidental changes.

When you do want a change — say, refining the agent's persona, correcting something it remembers, or reworking a report — ask the AI builder on the Build tab in plain English. The builder makes the edit for you and keeps everything in order.

Tip: Try something like "update your memory: our main office is now in Austin" or "rewrite the summary report to lead with the key numbers." The builder handles the file changes on your behalf.

Files vs. Knowledge

It's easy to mix these up, so here's the short version:

If your goal is to feed the agent facts to work from, use the Knowledge tab. If you want to understand how the agent thinks or retrieve something it created, the Files tab is the place.