In October 2025, the American AI laboratory Nof1 did something. Six large language models, each with $10,000, were thrown into the Hyperliquid exchange to trade cryptocurrencies on their own, without any human intervention.

DeepSeek V3.1 made a profit of 46%. GPT-5 lost 75%.

This competition was called Alpha Arena, which ran for two weeks, with all trading records publicly available on-chain.

It answered a question: Can AI trade cryptocurrencies?

The answer is yes. But it raised a bigger question: How can ordinary people participate? You can see how much DeepSeek made, but you can’t create an AI trader to compete with it.

Moss aims to solve this issue.

You tell it how to trade, and it helps you create an Agent

Moss launched an open platform (moss.site/agent).

The task is simple: you describe in plain language how you want to trade, and the AI converts this into a complete quantitative strategy, which is then deployed as an automatic trading Agent.

Here are a few examples. You say, "I want to do trend reversal," and it generates a trend reversal Agent. You say, "Long and short hedge steady as a dog," and it configures the parameters accordingly. You say, "Aggressive volatility hunter," and it creates a high-frequency, high-volatility strategy for you.

No coding skills are needed. No need to understand what moving averages, Bollinger Bands, or RSI are. Free of charge.

All you need is an OpenClaw or Claude Code environment. Open the terminal and enter a command:

clawhub install moss-trade-bot-factory

Then tell it how you want to trade, bind a pairing code, and your AI trader will be online. Two messages are all it takes.

Previously, if you wanted to run a quantitative strategy, you had to know Python, understand how to set parameters for technical indicators, and build your own backtesting framework. The threshold was high. Moss compresses this entire process into a single conversation.

Who is Moss

Before developing the AI Trading Agent, Moss already had an operational product. It is a Chrome browser extension that, once installed, integrates into your X (Twitter) page, providing real-time market summaries, KOL opinion aggregation, and on-chain Alpha signal tracking. In simple terms, it is an AI assistant for crypto information.

The AI Trading Agent platform is the latest module added to the Moss product line.

There are already quite a few AI tools in the information layer market, such as Kaito and various AI feed products. However, Moss may be among the first to allow users to create trading Agents with zero barriers and compete publicly.

Two modes: testing you with history, validating you in real-time

Once the Agent is created, there are two ways to run it.

The first is called Hell Mode. The platform took the real BTC market data from the 150 days starting from the major crash in October 2025. All Agents are thrown into the same historical trend to run. The starting point is the same, the data is the same, the only difference is the strategy.

Why choose this data? Because during these 150 days, there were all kinds of situations. Sharp declines, sideways movements, false breakouts, rebounds, and corrections. If a strategy can only make money in a unidirectional market, it will perform poorly in this dataset. Hell Mode tests the risk resilience of the strategy.

The second is called Live Mode. It connects to real-time market data, and every operation of your Agent, every change in position, and how much it earns or loses is updated in real-time.

In both modes, the PnL (profit and loss) leaderboard for all Agents is fully public. You can see your ranking and also what styles and performances others’ Agents have. Each mode has its own independent leaderboard.

Having an open leaderboard is very important. Every strategy is subject to scrutiny by everyone, with no black box. If you claim your strategy is powerful, you can prove it on the leaderboard.

The Agent learns on its own

Moss has a design detail worth mentioning separately.

Traditional quantitative strategies are static. Once the parameters are confirmed through backtesting, they remain largely unchanged until the strategy fails and is manually adjusted. During this time, if the market style changes while the strategy continues to use old parameters, it's highly likely to lose money.

Moss’s Agent has a weekly evolution mechanism. After each operation cycle, the Agent automatically adjusts its parameters based on its performance that week. If it loses too much, it minimizes risk by reducing position sizes and tightening stop losses. If it performs well, it increases the weight of the advantageous strategies within the risk management range.

This mechanism aims to simulate the behavior of a real trader. A good trader does not strictly stick to one set of parameters; instead, they adjust their strategy according to market conditions. Moss wants the AI Agent to have this capability as well.

The effectiveness will depend on the design quality of the underlying algorithm and its adaptability to different market environments. The 150-day Hell Mode data serves as a testing window, while longer-term validation will take time.

How to participate

Currently in the public testing phase, it is free, no wallet connection required, and no quantitative background needed.

Step one: Install Skill

In the OpenClaw or Claude Code environment, enter:

clawhub install moss-trade-bot-factory

Skill address: clawhub.ai/fei-moss/moss-trade-bot-factory

This Skill is the predefined strategy generation framework of the Moss platform, serving as the basic component for creating Agents.

Step two: Create Agent

Send a message to OpenClaw describing your trading style in natural language. It can be quite general, such as "Buy low and sell high in a volatile market, don't be too aggressive." Or you can be more specific, like telling it the maximum drawdown you're comfortable with or your preferred holding period. The AI generates strategy parameters based on your description and deploys them automatically.

Step three: Bind pairing code

Follow the prompts to bind the Agent with the Moss platform, and the Agent will start running in a simulated environment.

Step four: Check the leaderboard

All Agent leaderboard entry: moss.site/agent

Hell Mode and Live Mode have independent leaderboards where you can view returns, strategy descriptions, and operating status.

From installation to Agent going live, two messages. The author tested and installed their own strategy, resulting in a 37.47% ROI.

Future Plans

It is understood that the current version is the first phase, supporting the creation of standardized Agents using public Skills. More capabilities will be gradually opened up in the future.

First, open external data API access. Users can connect more signal sources to their Agents, not limited to the platform's default data.

Second, support for uploading custom strategy Skills. Users with a quantitative background can write and upload their own trading logic, allowing Agents to run based on their framework.

Third, launch Hosted Agent services. Users without OpenClaw or Claude Code environments can also create and run Agents directly on the platform.

As the Agents evolve to this stage, the AI Trading Agent landscape is rapidly developing its infrastructure.

On the payment side, x402 is expanding rapidly under the promotion of Coinbase and Cloudflare. As of October 2025, the protocol had processed over 520,000 transactions, and the developer community has incubated over 200 projects based on x402. These two numbers continue to grow.

There is beginning to be a differentiation at the application level. Nof1's Alpha Arena is a closed experiment, measuring which AI model has stronger trading capabilities. The open-source project AI-Trader on GitHub follows a signal market route, where Agents publish trading signals for others to follow. Moss chose a third path, creating an open platform that allows everyone to build their own AI traders and compete publicly.

Who can trade, whose signals are good, everyone can participate. Three directions, three different bets. Moss bets on the last one.

How far this path can go depends on two factors. First, whether the strategies generated from natural language can continue to make money in real markets. Second, once there are many users, whether the Agents created will become more similar, leading to a convergence of strategies and the disappearance of alpha. It's impossible to answer now; we will observe slowly once the leaderboard is up and running.

---

Rhythm BlockBeats is looking for more AI Agent Trading products, feel free to recommend.

We can bring you maximum exposure and early users, feel free to DM us on Twitter @BlockBeatsAsia

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。