🧞Today & Tomorrow of AI Agents with Crestal Network

AI Agents today and tomorrow!

Mar 10, 2025

Gm and happy Monday frens🖖

AI agents, AI Agents...

We've been talking about them since the end of the last year and a whole January. We've seen many (and I want to stress "many"!) degen projects something similar to memecoins.

However, we (and hopefully you!) believe in AI Agents. What is the currents state and what are the main problems - further:
🦾 AI Nation stand up
🧵 ETH speed and DeFi lies
📰 News

🔮 You're going to order merch here...

Merch is one of the most important parts of a successful event: people will wear it and remember your project long after it's over.

That's what we do in-house, in addition to organising great events: our DeHugs team designs and produces cool merch to keep you on top of mind.

Get cool merch for your next event!

Contact via Telegram

Contact via Email

🛠️ Deep-dive: AI Agents today and tomorrow!

Crestal recently pivoted from modular development platform for Dapps toward AI agent development. They repositioned their platform to enable users to create and deploy AI agents with customized functionality. These AI agents could be tailored for various purposes, such as:

DeFi trading bots that automate asset management, yield farming, and portfolio rebalancing.
Social engagement agents that manage community interactions, schedule posts, and curate content.
Research and data analysis agents that scan blockchain data, news, and sentiment to generate actionable insights.

Insights from a new podcast with Marouen Zelleg, Co-founder @ Crestal Network. Watch the full episode here.

The AI Agent ecosystem

The AI agent ecosystem is evolving fast, with platforms taking different approaches to building and deploying AI-driven automation.

Some platforms, like Eliza, focus on providing a powerful AI agent framework for experienced developers, allowing them to build and refine agents with high precision. However, these solutions are not accessible to non-technical users.

On the other end of the spectrum, platforms like Virtuals and Pump.fund focus on launching AI agent tokens rather than functional AI agents. While these projects generate hype, they often lack meaningful use cases beyond token speculation.

Challenges & risk for AI Agents in DeFi

The adoption of AI agents in DeFi faces two major obstacles: trust and reliability. Users don't want to hand over financial control to AI due to risks like misinterpreted commands, security vulnerabilities, and unexpected market conditions.

One of the main risks is execution reliability - ensuring that an AI agent understands and performs the intended action correctly. Without proper safeguards, an agent might make poor trading decisions, leading to losses.

Security is another major concern, as malicious actors could manipulate AI agents to perform unauthorized transactions. Furthermore, if an agent interacts with bridging protocols, the risk of transaction failures increases.

Crestal addresses these concerns by implementing controlled environments for AI agents. Instead of giving an agent full access to financial markets, Crestal restricts the assets, liquidity pools, and chains it can interact with. This approach ensures that an AI cannot accidentally drain a wallet or execute risky transactions beyond predefined limits.

The platform also relies on incremental autonomy—agents start with limited permissions and gradually expand their capabilities as they demonstrate reliability. This step-by-step approach ensures that AI-driven financial decisions remain under control. Additionally, Crestal promotes a human-in-the-loop design, where users fine-tune their AI’s behavior through prompt adjustments rather than allowing complete autonomy from the outset.

Monetization for AI Agents

Finding a sustainable revenue model is one of the biggest challenges for AI agent platforms. Crestal is exploring two primary approaches: pay-as-you-go pricing and revenue sharing with developers.

Under the first model, users pay a fixed fee to create an AI agent and an additional fee for each transaction it executes. This approach is similar to traditional SaaS pricing, ensuring that users only pay for what they use. It also helps eliminate speculative behaviors tied to token-based AI agent models.

The second model focuses on developer monetization. Developers who build custom AI agent skills—such as trading strategies, social engagement algorithms, or data analytics modules—can earn a share of the revenue whenever their modules are used. This approach turns Crestal into a marketplace for AI capabilities, where skilled developers contribute to an expanding ecosystem while being compensated for their work.

Role of blockchain in AI Agents

Blockchain plays a critical role in AI agents, but its implementation comes with unique challenges. Many assume that AI agents must exist on-chain, but in reality, AI agents function off-chain while interacting with blockchain-based applications.

The primary role of blockchain in AI agents is payments and incentive structures. AI agents need a transparent and secure way to handle transactions, whether for trading, subscriptions, or service fees. Smart contracts ensure that AI-driven actions are verifiable and immutable.

Interoperability is another key function. Since AI agents operate across multiple blockchain ecosystems, they must be able to interact seamlessly with different chains. For example, an AI agent optimizing yield farming should be able to compare APYs across Ethereum, Base, and Solana and decide where to allocate funds. However, cross-chain execution introduces complexity, as agents must handle bridging risks, slippage, and liquidity variations across different networks.

Another major challenge is privacy and security. AI agents process sensitive data, and if all interactions were on-chain, they could be exposed to potential exploits. Zero-knowledge AI solutions may become necessary in the future to protect user data while maintaining transparency.

Listen to the full episode on all platforms:

Spotify

YouTube

Apple Podcasts