Mantis: The Platform for DeFAI

“DeFAI” is a trending term - but it represents what we have always been doing with intents, execution, and settlement on Mantis - there is just now a term for it.

In the DeFi space, AI agents are no longer speculative concepts—they are actively participating in and influencing financial ecosystems. However, there is a lack of secure and reliable infrastructure for AI agents to execute their strategies effectively. Mantis fulfills this market need to provide a platform for DeFAI, as described below.

We’ve already built the execution platform, which is up and running. We’re now working on increasing our AI-specific products and tools, starting with a large language model (LLM) and then expanding from there.

Keep reading to find out how we’ve been focusing on DeFAI since the start of Mantis, and how we’ll keep ramping up our AI support in the future:

It’s Always Been DeFAI at Mantis

The term “DeFAI” has taken CT by storm, putting words to the premise of using AI to optimize user interactions and outcomes within the DeFi space. But, at Mantis, we’ve always been focused on allowing the user to input their desired outcome, with optimal routing and execution carried out on the backend. In fact, this infrastructure is already built and in use.

This process of abstracting away the actual route selection and execution from users has only recently come to be known as DeFAI: users state what they want accomplished in the DeFi space, and agents carry this out for them. We have approached this via intents and our solver network, which were created to allow the optimization, routing, and execution of various DeFi actions (currently, swaps).

We’ve been talking about these ideas for a long time before the term DeFAI arose, such as in the Mantis whitepaper from May 2024. For example in the last line of our abstract, we state our goal is to enable “expressing preferences across blockchain domains”.

As another example, here is a recording of our presentation at Breakpoint where we demoed the Mantis app. In this presentation, Mantis Co-Founder and Research Director Omar Zaki (@0xbrainjar) described how Mantis powers best execution for users’ desired transactions/intents.

Since we have been working on this goal for some time, unlike newcomers to the DeFAI space, we are already in production. Now, our expression and execution platform is live, replete with swaps layer and routing layer, and is processing AI-generated intents (see below).

AI Agents, Live Today on Mantis

We have a number of AI trading bots active on Mantis. This includes Lady Mantis, who publicly tweets about her trades. In the future, we will continue creating even more advanced AI agents on Mantis. In fact, Mantis envisions intelligent intents to become the mechanism with which AI interacts with the blockchain. We believe that, in the future, all users will have personalized AI agents with which they operate on-chain - which we intend for Mantis to provide. For example, you can tell your AI agent on Mantis “buy X SOL if it drops to Y price” or “sell memecoin Z when I’ve gone 5x”. From here, you won’t need to give it another thought - your own AI agent manages the rest, optimally executing your preferences.

The Future of AI Trading Infrastructure

We are working to further increase our products and infrastructure that are uniquely catered to AI agents. This includes the Mantis LLM (large language model). LLMs are a type of AI that can process plain text/natural language. Our LLM will function to take in written requests from users to execute crypto tasks on Mantis. This will make it even easier for users to interact with DeFi, using language of their own choosing. Keep an eye out for more info on this coming soon.