TinyAGI Developer Guide
TinyAGI is an experimental multi-agent, multi-team, multi-channel assistant runtime with a browser control surface called TinyOffice. This guide explains what TinyAGI is, how to get it running, how the queue and team model fit together, and where to extend it.
This guide is based on the official TinyAGI/tinyagi repository, checked on June 26, 2026.
In this section​
- Setup and first run
- Architecture and runtime
- Configuration and security
- Extending and contributing
- Simple user guide
1. What TinyAGI is​
The shortest accurate description is:
TinyAGI is a multi-agent runtime where isolated agents, teams, chat rooms, and messaging channels work together under one local control plane.
The README explicitly highlights:
- multi-agent,
- multi-team,
- multi-channel,
- persistent sessions,
- a SQLite-backed queue,
- and a web portal called TinyOffice.
2. The mental model​
Think of TinyAGI as five cooperating layers:
| Layer | What it does |
|---|---|
| agent runtime | Runs isolated agents with specialized roles |
| teams and chat rooms | Coordinate collaboration between agents |
| channel adapters | Connect Discord, Telegram, WhatsApp, and more |
| SQLite queue and persistence | Keep jobs, retries, and sessions durable |
| TinyOffice portal | Gives operators a browser-based control plane |
3. Why developers would choose it​
- It is explicitly multi-agent instead of single-assistant-first.
- It supports team collaboration and fan-out patterns.
- It includes both CLI/TUI and web portal surfaces.
- It is built to run 24/7 as a background process or Docker container.
4. What makes it distinctive​
TinyAGI is not trying to be the smallest possible coding CLI. It is aiming at an always-on coordination system where multiple agents and teams can keep working across channels and tasks.
5. Recommended reading order​
- this overview,
- quick start,
- TinyOffice and channels,
- architecture and queue design,
- plugin and extension surfaces.