The Almanak Strategy Builder: How to create your own strategy with the help of an AI swarm

Using AI swarms to build trading strategies

Recently, AI agents have returned to the spotlight for two main reasons.

First and foremost, x402, an open-source protocol developed by Coinbase that enables automatic micropayments and then we had the launch of Alpha Arena, an interesting experiment which gave $10k to the six most prominent generalised LLMs and sought to analyse their trading performance.

Alpha Arena aimed to answer the question: Can AI agents actually make money?

We’ve dived deep into it in one of our previous reports, highlighting the limitations of LLMs and how others, such as Almanak, are developing frameworks and swarms of agents tailored to this task.

While agents can help iterate and make strategies executable, the human element remains central to the Almanak design.

The Almanak platform is finally open for all users who applied to the whitelist, with open access to anyone set for November 27th.

In this article, we dive into the Almanak’s Strategy Builder, the part where users can create and implement strategies, helped by a swarm of agents. We briefly introduce the protocol’s basic functioning, then test it by creating and implementing a Pendle RSI (Relative Strength Index) strategy.

From General Frameworks to Specific Implementation

LLMs can assist users with a variety of tasks, such as depositing funds into a specific vault, purchasing a token, or making everyday transactions.

However, when it comes to executing complex strategies, LLMs require extensive instructions, and the primary issue is that the final execution is neither verifiable nor deterministic, but rather probabilistic. Furthermore, generalised LLMs also lack precision concerning financial investments.

Currently, implementations also struggle to execute complex strategies and tend to focus more on simple operations. Last but not least, LLMs can often hallucinate and are open to attacks such as prompt injection. While agents act independently without human intervention, this also results in a lack of accountability.

All in all, these factors make it hard for these strategies to scale, as institutions are reluctant to have their funds exposed to these risks.

For those interested in diving deeper, we extensively discussed this topic in our previous report.

To solve these issues, Almanak has a different philosophy.

Agents are leveraged for what they do best: creating and iterating on strategies. However, strategies are hard-coded rather than executed by AI agents.

As part of this, users can create new strategies in the Almanak’s Strategy Builder.

Cooking up Strategies in the Almanak Strategy Builder

As the name hints, this is where magic happens, where strategies are actually crafted and implemented.

Users can select any available template and gain insight into the strategies implemented by watching the demos, with custom strategies coming soon.

It’s essential to note that, given the central role of human intervention, users must possess moderate to advanced knowledge of the underlying strategies to fully leverage Almanak’s agent swarm.

In fact, these are only templates to be interpreted as guides for users (called Builders on Almanak) to improve and iterate on.

There are two layers of Almanak that complete the strategy from ideas to execution:

1. Creation Layer: This is where the strategy is designed, simulated and iterated with the help of an AI agent swarm.

2. Execution Layer: When the strategy is ready, users can deploy it onchain through this layer, ensuring secure, permissioned, and verifiable deployments.

There are several technical trading strategy templates currently available, such as:

These strategies are presently executable on Ethereum, Arbitrum and Base.

Here is a practical example of a strategy that you can create and leverage in the Almanak Strategy Builder. In this case, we used the demo template to create an RSI strategy on Pendle.

Pendle RSI Strategy

To create this strategy, we used the available template for the Pendle RSI Momentum strategy.

After selecting the strategy, you will be redirected to the technical analysis strategy page, where you can customise all aspects and preview your strategy.

For this article, we replicated the template strategy without any changes. Still, it’s important to highlight that these strategies provide extensive customisation as users could change variables such as:

• Chain: Arbitrum, Base, and Ethereum are currently available.

• The token of the strategy: In this case, it’s PENDLE, but you can change to BTC, ETH or others.

• Signal Parameters: Adjust the strategy type and the indicator levels; in our case, define the RSI value at which the Strategy should buy or sell the token.

• Execution Risk: Determine the cooldown period between trades and the size of your portfolio that you wish to trade with.

Once you are satisfied with the parameters, you can create your own custom strategy based on them. When you create this strategy, Almanak’s agent swarms write, debug, and review the code.

Upon creating the strategy, it will be available in your custom strategies, where you can deploy it onchain and use it to buy and sell the token of your choice when specific technical indicators are met.

Closing Thoughts

Almanak’s swarm is pushing the boundaries of agentic interaction with a specific focus on developing profitable trading strategies.

So, by asking whether agents can actually make money, we’re probably asking the wrong question. Rather, we should be asking, can agents help you craft strategies that will potentially make you money?

For this is really where they come into play, at least within the Almanak framework. However, the human figure remains central, which means these strategies are targeted at users who have at least some basic knowledge of them.

As we mentioned earlier, all the users who were on the waitlist can now access the Almanak strategy builder with some test credits and access two types of strategies:

• Technical Analysis (TA) based strategies, for example, the Pendle RSI Momentum Strategy

• Liquidity Provisioning (LP) strategies on Uniswap V3

Moreover, if you were not on the waitlist, the platform is going to be open for everyone by the end of this month.

However, even if you are not an expert, you can still experiment with the available strategies and learn as you go.

As agents become increasingly present within our daily realities, becoming able to leverage them efficiently is not just a possibility, but a prerequisite in order to stay ahead of the curve and ensure to continue having a hedge when these tools become ubiquitous.

We hope you enjoyed this article and give the strategy builder a try!