Welcome to the Vertex Swarm Challenge 2026. This hackathon pushes the boundaries of how we think about coordination. In a world where billions of devices—from warehouse robots to autonomous drones and AI agents—need to work together in real-time, the traditional "cloud-first" model is hitting a wall.
This guide will help you understand the core concepts of Swarm Coordination, the tools at your disposal (Vertex and FoxMQ), and how to choose the right architecture for your project.
At its core, Consensus is the process by which a distributed network of participants (nodes) agrees on a single version of the truth, even if some participants are faulty or malicious. In the context of a robotic swarm, consensus means every robot agrees on the exact order of events or movements without needing a central director.
A Swarm is a decentralized collective of autonomous agents that interact locally with one another and their environment. While individual agents may be simple, the collective behavior emerges as complex, coordinated action. Think of a flock of birds or a colony of ants—there is no "leader," yet they move as one.
| Feature | Centralized (Cloud-First) | Swarm (Vertex-Powered) |
|---|---|---|
| Control | Single Controller / Cloud Server | Distributed / Peer-to-Peer |
| Connectivity | Star Topology (All nodes to center) | Mesh Topology (Node to Node) |
| Latency | Dependent on Round-trip to Cloud | Edge-local (<100ms) |
| Failure | Single Point of Failure (SPOF) | Byzantine Fault Tolerant (BFT) |
| Scalability | Bottlenecked by Central Capacity | Grows with the Network |
- Byzantine Fault Tolerance (BFT): In a centralized system, if the controller sends a corrupt command or is hacked, the whole system fails. In a Swarm using Vertex, the network can reach agreement even if 1/3 of the nodes are behaving erratically or maliciously. This is critical for safety-critical hardware like drones.
- Fair Ordering: Traditional central queues (like RabbitMQ) order messages based on when they hit the server. In a high-speed swarm, this isn't "fair" because network jitter can change the perceived order. Swarms use consensus to ensure that the actual sequence of events is mathematically preserved across all nodes.
- Gasless Edge Processing: Unlike blockchains which require "gas fees" for every transaction, Vertex allows for high-throughput coordination at the edge. A drone swarm making 1,000 adjustments per second can't wait for a block confirmation or pay a fee for every wing-flap. Swarms process the work where it happens.
- Resilience to "Cloud Rip": A DNS outage or a cloud region failure can ground an entire fleet of centralized robots. A swarm operates as a "Meshnet"—as long as the robots can see each other (via WiFi, LoRa, or 5G), they can keep working.
Vertex is Tashi’s embedded consensus engine. It is the "brain" that allows individual nodes in a swarm to reach agreement in sub-100ms.
- Powered by Hashgraph: Vertex uses a Directed Acyclic Graph (DAG) and "Virtual Voting" to achieve consensus. Unlike traditional blockchains, it never needs to send "vote messages" across the network, making it incredibly fast and efficient.
- Meshnet Architecture: When you use Vertex, you aren't just building an app; you're creating a peer-to-peer network where participants are both users and consensus nodes.
- Private by Default: Only the final "Proof of Coordination" needs to leave the local swarm. Internal state and intermediate consensus rounds stay within your meshnet.
FoxMQ is a decentralized MQTT broker built on top of the Vertex consensus engine. If you are already familiar with IoT development, FoxMQ is your bridge to the swarm.
- Familiar Protocol: It speaks MQTT—the industry standard for IoT—meaning your existing robots and sensors can connect to it immediately.
- Consensus-Backed Queues: Unlike a standard broker that lives on one server, FoxMQ distributes the queue across the entire swarm. If one broker node goes down, the others don't skip a beat.
- Fair Broadcast: FoxMQ ensures that every subscriber receives messages in the exact same order, which is vital for synchronized maneuvers in a drone fleet.
Choosing your tool depends on how "deep" you want to go into the consensus logic.
- You are building Custom Logic: You need specific rules for how robots move or how agents negotiate.
- You need Maximum Throughput: You are writing native Rust or C code for high-performance edge devices.
- You want Native Consensus: You want to embed agreement directly into your application's state machine.
- You have an Existing Fleet: Your devices already use MQTT and you want to "decentralize" them without rewriting their firmware.
- You need Simple Pub/Sub: Your main goal is sending messages (e.g., sensor data, health pings) across many parties.
- You want Low Friction: You need to get up and running quickly with standard MQTT libraries.
Get started with the following docs and tools:
- FoxMQ Python Warmup: Get started with decentralized MQTT using Python.
- Vertex Rust Warmup: A template for building native consensus-powered agents in Rust.
- Webots Starter Kit: Orchestrate robotic simulations in Webots using Vertex.
- AirSim Vertex Starter Kit: Decentralized drone swarm coordination starter kit using Tashi Vertex P2P consensus and Microsoft AirSim.
- Nexus Agentic Commerce: High-Frequency Agentic Commerce built on Tashi Vertex.
- Pioneer Swarm Exploration: Comprehensive ROS 2 + Webots stack for coordinating five Pioneer robots that cooperatively explore an arena, share telemetry through FoxMQ, and reach consensus on robot health before accepting detections.
- Tashi Vertex Rust Context: Get up-to-date documentation and code context for Tashi Vertex RS optimized for LLMs and AI agents via Context7.
- FoxMQ Agent Skill: A specialized skill definition for FoxMQ. Add this file to your repository's
.agent/skills/directory (or runnpx ctx7 skills install /tashigit/foxmq foxmq-mqtt-broker) to give your AI agent deep context on FoxMQ cluster setup, producer-consumer patterns, and consensus-backed MQTT networking.
- Tashi Documentation: The home for all things Vertex and FoxMQ.
- Vertex for Rust (vertex-rs): Build native Rust-based swarm agents.
- Vertex for C (vertex-c): Integrate consensus into C
- FoxMQ Resources: Deep dive into the decentralized MQTT broker.
- Technical Whitepaper: For those who want to understand the DAG and virtual voting math.
- Tashi Glossary: Catch up on the terminology.
Tashi is the coordination layer for intelligent systems. We provide the infrastructure that allows autonomous machines—robots, drones, and AI agents—to reach agreement in milliseconds without needing to trust a central authority.
Our mission is to align devices with the interests of their owners. Traditionally, devices serve the platforms they are connected to; Tashi flips the script by enabling coordination at the edge, where machines operate with verifiable alignment to those they represent.
The Tashi stack consists of three powerful layers:
- Vertex: The consensus layer (used in this hackathon) providing sub-100ms BFT agreement.
- Lattice: A global DePIN infrastructure layer for node discovery, validation, and economic rewards.
- Arc: A settlement bridge that connects private meshnet coordination to public blockchains for payments or public finality.
Tashi is also a co-founding member of the Intercognitive Foundation, working alongside pioneers like Auki and Peaq to establish global standards for AI accessibility and the machine economy.
Good luck, Hackers! Build something that moves the world—literally.
