Skip to content
TinyFat Docs

What is TinyFat?

TinyFat gives you a personal AI agent that runs in the cloud. You email it, it thinks, it replies. No apps to install, no API keys to manage — just email.

The core idea

Section titled “The core idea”

Your agent runs in a secure container powered by pi-mono, an open-source agent framework. It has:

  • Its own email address — [email protected]
  • Persistent storage — Files survive between conversations
  • Memory — A MEMORY.md file that persists your agent’s context
  • Skills — Pre-installed tools for web search, file management, and more
  • Full shell access — Run any Linux command

How it works

Section titled “How it works”
  1. You email your agent — Send to [email protected]
  2. Agent wakes up — Container starts automatically
  3. Agent thinks — Reads your email, uses tools, reasons
  4. Agent responds — Writes a reply to its outbox
  5. You get the email — Platform delivers the response

Each email triggers one agent “turn” — the agent processes your message, does its work, and responds. This is simpler and more reliable than long-running processes.

What makes TinyFat different?

Section titled “What makes TinyFat different?”

No setup required

Section titled “No setup required”

Sign up with your email, get $1 free credits, start chatting. No API keys needed to get started.

Transparent pricing

Section titled “Transparent pricing”
  • Free tier: $1 credits to try it out
  • Tiny tier ($5/mo): Custom email, Slack, metered API at cost
  • BYOK (coming soon): Bring your own Anthropic key, small platform fee for email infrastructure

We never mark up API costs.

Radical transparency

Section titled “Radical transparency”

We don’t hide the implementation. Your agent runs pi-mono in a Cloudflare container. The code is open source. You can inspect everything.

Escape hatch

Section titled “Escape hatch”

Want to run this yourself? Everything is open source. Export your data anytime. We’ll even tell you how to set it up on your own infrastructure.

What can your agent do?

Section titled “What can your agent do?”

Out of the box:

  • Answer questions using Claude’s knowledge
  • Browse the web with Brave Search
  • Read and write files in its workspace
  • Run shell commands for automation
  • Remember context across conversations
  • Generate PDFs with Python

With additional setup:

  • Post to Slack (Tiny tier)
  • More integrations as you add them

Email commands

Section titled “Email commands”

Control your agent with special commands:

CommandEffect
/clearStart a fresh session
/resumeList recent sessions
/resume abc1Resume a specific session
/compactCompress context to save tokens
/modelShow current model
/model haikuSwitch to Haiku (fastest)
/model sonnetSwitch to Sonnet (balanced)

Next steps

Section titled “Next steps”