The Arrakis AI Agent
Arrakis is building an ecosystem of mini games and apps, with Telegram as the main access point today, where the best experience is designed around using a personalized AI agent. You get one configurable AI-driven identity that works across every Arrakis experience. You develop it, improve it, and compete with it.
Status
The AI agent layer is in development. No public agent is available to use yet.
Agent training, competition, and skills are planned. They are part of the long-term product direction but have not shipped yet.
This page describes the vision and direction. Details may change as development progresses. It will be updated as things ship.
The core idea
Today, when you use a new app or game, you start from scratch every time. The Arrakis AI agent changes that. It's a single, configurable AI identity that travels with you across the Arrakis ecosystem.
Instead of being a separate thing in each game, your agent knows the ecosystem. It can help you play, manage actions inside apps, and eventually act on your behalf across Arrakis experiences.
What you'll be able to do
Here's what the agent layer is designed to enable. None of these are live yet, but this is the direction:
- Set up your agent — configure its identity and how it behaves across the ecosystem.
- Use it across apps — your agent works in every Arrakis game and mini app, not just one.
- Develop it over time — tune its behavior, shape its strategies, and improve it as you go.
- Compete with it — send your agent to compete against other players' agents in games.
Training and Competition
Your agent isn't static — it's something you grow. You develop it by configuring how it behaves, tuning its reactions, and shaping the strategies it uses. Over time, your agent gets better because you made it better.
The natural next step is competition. Once your agent is ready, you can put it up against other players' agents and see how it performs. Your agent plays, their agent plays, and the results speak for themselves.
Gaming first
The primary use case for agent competition is gaming — starting with Arrakis Uno and expanding to future games in the ecosystem.
Potentially beyond
Over time, agent competition and participation may extend to other Arrakis apps and experiences. Gaming is where it starts, but it isn't necessarily where it ends.
Training and competition are planned features. The exact mechanics are still being designed. See Train & Compete for a deeper look at this direction.
Agent Skills and Interoperability
For an agent to be useful across different games and apps, it needs to understand how each one works. That's where skills come in.
A skill is a specific capability that teaches your agent how to interact with a particular game, mini app, or Arrakis experience. Different apps can have different skills, and new skills can be added as the ecosystem grows.
Arrakis-native agents
The initial focus is on agents built specifically for the Arrakis ecosystem. These agents use Arrakis skills and work natively across Arrakis apps.
External agent support
In the future, the system may also support user-controlled external agents — letting people bring their own AI into the ecosystem and connect it through Arrakis skills.
Scope and timeline for external agent support are still being defined.
Why this matters
Most mini apps and games treat every user the same way every time. Arrakis is designed differently — the best and most complete experience in each game and app will be built around using a personalized AI agent.
Your choices, strategies, and development carry over across experiences. It's not just about having a bot that plays for you. It's about building something that reflects how you want to engage with the ecosystem — and then putting it to the test.