I built a 17-agent AI swarm on my phone — here's how

Description and use

A 17‑agent AI swarm runs entirely on a phone, leveraging Godot, TinyLLM, and local ChromaDB for inference and memory.

Problem

Run a swarm of 17 independent AI agents entirely on a mobile device without cloud inference, to explore edge AI and emergent behavior while preserving privacy.

Outcome

A 17‑agent swarm operates entirely on a smartphone, with local inference and memory, demonstrating feasible edge AI swarm behavior.

Workflow

• 1. Chose Godot Engine as the runtime platform.
• 2. Implemented agents in GDScript, each with a unique prompt and role.
• 3. Integrated TinyLLM (1.3B 4‑bit quantized) for local language inference.
• 4. Set up a local Flask API to expose TinyLLM and ChromaDB services.
• 5. Each agent communicates with the Flask API over HTTP to query or update the vector database.
• 6. Agents loop through their tasks, exchanging observations, plans, and feedback.
• 7. The system runs continuously on the phone, demonstrating emergent swarm behavior.

Evidence

running a swarm of 17 independent AI agents entirely on my phone… everything is local, leveraging the surprisingly capable hardware we carry around in our pockets.

Source

TinyLLM is a quantized language model that can run on CPUs, even relatively low-powered ones. I’m using a 1.3B parameter model, quantized down to 4-bit, making it significantly smaller and faster.

Source

Each agent maintains a short-term memory using a local ChromaDB instance.

Source

The Flask server receives the questions, queries the ChromaDB vector database using TinyLLM for semantic similarity search, and returns the most relevant answer.

Source