Hello everyone!
I’d like to share with you Automata, a Registry and Discovery System designed specifically for multi-agent ecosystems.
For the next step in the future of the Internet of Things and autonomous agents, I believe a project like this is essential.
What problem do we solve?
Today, when an LLM needs to use a tool (an Agent/API), the list of available tools is usually either hardcoded or static.
- Discovery is Manual: If your LLM needs a flight-quote service, you must know exactly which API to use and integrate it manually.
- Lack of Quality: There are no performance metrics (latency, success rate) to help the LLM choose the best service.
- Lack of Scale: Maintaining hundreds of external services up to date is nearly impossible.
The Automata Solution: Intent-Based Routing
Automata works like DNS for AI Agents.
Service Agents (Providers) register themselves, and Consumer Agents (your LLM) search for the service they need based on user Intent, not on an API name.
Example of Automatic Discovery
// The Consumer (Your LLM) searches for the right Agent based on Intent
const consumer = new AgentConsumer({ llm: { ... } });
// LLM analyzes the user request: "Where can I get the best price for lettuce in São Paulo?"
const agents = await consumer.search({
intent: 'food.supermarket.estimate', // Intent (Rigid Structure: DOMAIN.ENTITY.ROUTE)
location: 'São Paulo, Brazil',
tags: ['fresh-produce', 'delivery'],
});
// Execution: Automata returns 3–5 Agents ranked by performance, success, and geographic proximity.
const quotes = await consumer.executeMultipleWithFeedback(agents.slice(0, 3), {
task: 'get_quote',
params: { item: 'lettuce' }
});
Intelligent, Merit-Based Ranking
Agents aren’t ranked only by metadata, but by real performance.
Automata uses 9 factors in its ranking algorithm, giving strong weight to Intent match, location, and success/latency metrics.
| Factor | Weight | Optimization (AIO) |
|---|---|---|
| Intent Match | 25% | Use domain.entity.action format (e.g., finance.loan.estimate) |
| Geographic | 20% | Define precise geographic scope |
| Success (Feedback) | 14% | Maintain a high success rate |
| Categories/Tags | 17% | Use specific categories & tags (AIO!) |
| Latency/Rating/Fraud | 24% | Deliver speed and quality |
Key Concept: AIO (Agent Intent Optimization)
If SEO optimizes for Google, AIO optimizes for Automata.
We teach LLMs to extract Intents using the structure DOMAIN.ENTITY.ROUTE.
Your Provider Agent must register with precise Intents and Categories to maximize discovery.
Example of strong AIO:
intents: ['food.restaurant.search', 'food.delivery.search'],
categories: ['cuisine:italian', 'place-type:trattoria'],
tags: ['pet-friendly', 'cash-only', 'delivery-sp'],
What you can do with Automata
• Service Provider: Register your API/function once and be discoverable by any LLM or Agent in the ecosystem (zero-config SDK).
• Service Consumer: Give your LLM the ability to dynamically search and orchestrate services, without hardcoding.
• Orchestration: Use ranking to run simultaneous quotes and automatically choose the best provider.
Check out our GitHub and get started in 5 minutes:
We welcome feedback and discussions on how we can push tool discovery to the next level.