Skip to content
TinyFat Docs

Agent Workspace

Your agent has a persistent filesystem at /data. Files here survive container restarts.

Directory structure

Section titled “Directory structure”
/data
├── MEMORY.md           # Agent's persistent memory
├── context.jsonl       # Conversation history
├── CEMAIL/             # Email channel data
├── outbox/
│   ├── email/          # Pending emails to send
│   ├── sent/           # Successfully sent emails
│   └── failed/         # Failed email attempts
└── skills/
    ├── pi-skills/      # Core skills from pi-mono
    └── tiny-skills/    # TinyFat platform skills

MEMORY.md

Section titled “MEMORY.md”

This is your agent’s brain. It persists across sessions and contains:

  • Agent identity and configuration
  • Owner information
  • Instructions for the outbox pattern
  • Working notes and session logs

Editing MEMORY.md

Section titled “Editing MEMORY.md”

Your agent can update its own memory:

"Remember that I prefer bullet points over paragraphs."

You can also edit it directly through the dashboard’s file editor.

What to put in MEMORY.md

Section titled “What to put in MEMORY.md”
  • Preferences and working style
  • Project context and goals
  • Important dates or deadlines
  • Links to frequently accessed resources
  • Notes from previous sessions

The outbox directory

Section titled “The outbox directory”

Your agent sends emails by writing JSON files to /data/outbox/email/. See The Outbox Pattern for details.

Skills directory

Section titled “Skills directory”

Pre-installed skills live in /data/skills/:

pi-skills

Section titled “pi-skills”

Core skills from the pi-mono project:

  • Browser automation
  • Google CLIs (Gmail, Calendar, Drive)
  • Brave search
  • Beads issue tracking

tiny-skills

Section titled “tiny-skills”

TinyFat platform skills:

  • Firecrawl web scraping
  • Linear issue tracking
  • Notion integration
  • Slack API

Context file

Section titled “Context file”

context.jsonl stores conversation history in JSON Lines format. Each line is a message or tool result.

The agent uses this to maintain context across the session. It’s automatically managed by pi-mono.

File persistence

Section titled “File persistence”

All files in /data persist across:

  • Container restarts
  • Sleep/wake cycles
  • Platform updates

Files outside /data (like /tmp or /app) are ephemeral.

Accessing files

Section titled “Accessing files”

Dashboard file editor

Section titled “Dashboard file editor”
  1. Go to your dashboard
  2. Click on files in the sidebar
  3. View and edit directly in the browser

Via email

Section titled “Via email”

Ask your agent to read or write files:

"Create a file called projects.md listing my current projects."
"What's in MEMORY.md?"
"Show me the contents of /data/skills/tiny-skills/README.md"

Via the API

Section titled “Via the API”

Files are accessible through the dashboard API (authenticated).

Storage limits

Section titled “Storage limits”

Each agent has a 1GB volume. This is plenty for:

  • Text files and documents
  • Skills and scripts
  • Conversation history

Not recommended:

  • Large binary files
  • Media storage
  • Databases (use external services)

Next steps

Section titled “Next steps”