This week, many Web3 projects are integrating DeepSeek, with the fastest actions coming from Sora Labs, MyShell, Intent Kit, Regent, AlphaArc, Build fun, Autonome, Eliza, Hyperbolic, Eternal AI.
Among so many projects, MyShell supported DeepSeek V3 and integrated DeepSeek R1 in just 4 hours, becoming the fastest AI project in the BSC ecosystem to provide support and response for DeepSeek models. Their team's outstanding speed has been recognized by CZ.
1. Why choose MyShell for the demonstration
After completing the practical tutorial (only 3 steps, deploying DeepSeek locally), I originally intended to write another tutorial (using DeepSeek + Dexscreener to create a free AI trading robot), but this requires the deployer to be proficient in code editors like Visual Studio / VS Code or NOTEPAD++, which raises the overall threshold.
Therefore, I have been looking for no-code platforms to create my own AI Agent by combining no-code platforms with DeepSeek, enabling most people to get started.
MyShell is an AI Agent creation platform invested by YZiLabs, where users can create any AI application with zero code through the no-code AI development framework ShellAgent. Additionally, the ShellAgent development framework is equipped with three development modes (basic/professional/no-code) to suit different developers, and provides computing resources, model hosting, API calls, service orchestration, payment, and settlement infrastructure to maximize the development and operation of single AI Agent applications.
After integrating DeepSeek, MyShell allows users, regardless of their programming background, to create their own AI Agent according to the deployment guide.
In addition to DeepSeek and Flux models, the MyShell platform offers over 300 AI models as widgets, including GPTs, Claude, Janus Pro, Stable Diffusion, etc. Users can also use these models to create their own applications based on their ideas at shellagent.myshell.ai.
2. About Shell Coin
Since 2023, the MyShell project has been providing basic services for AI Agent creation. Currently, the project has accumulated 5 million users, 200,000 creators, and 20,000 AI Agent applications, with some creators' monthly income reaching $4,000, demonstrating a certain commercial monetization model and sustainability.
The official white paper shows that users can obtain $shell in four ways:
Creators need a Creator Pass, with the current floor price at 0.7E.
Holder of the Patron Badge;
Active participants in community incentives, test nets, and other related activities;
AI model contributors;
Providing products/services;
Currently, the fastest way to obtain $shell is the fourth method - AI model contributor. This time, MyShell integrating DeepSeek is just the opportunity for users to acquire $shell.
The official Twitter also stated: Build agents according to the tutorial, share experiences on X, and post links in our Discord. The first 50 shells will each receive 50 Shell Coins.
https://x.com/myshell_ai/status/1886762286737739785
So let's see how to participate:
3. Building AI Agent Tutorial
Since the official tutorial is already very detailed, I am using [official tutorial interpretation] for 'using templates'.
Step 1: Enter Shellagen
Related link: https://shellagent.myshell.ai/app, register
Click 'Create'
Select 'Create from Template' to create an agent using a template;
After clicking 'Create from Template', there are 9 templates available. The official website uses the first two templates as examples - 'X Agent-AI Child Book' and 'MyShell Agent-AI Child Book', each with different purposes.
In the official tutorial, it first introduces how to use the 'MyShell Agent-AI Child Book' template, followed by how to use 'X Agent-AI Child Book'. Don't confuse the two; if you're interested, you can create both in order.
After selecting 'MyShell Agent-AI Child Book', create a name in the dialog box, then click 'Confirm';
Then a ShellAgent workflow interface will appear, which contains a pre-built workflow using DeepSeek and Flux models (highlighted in red). In addition to these two models, it also offers over 300 AI models as widgets, including GPTs, Claude, Janus Pro, Stable Diffusion, etc. (highlighted in green).
Once the setup is complete, click 'Run' to test it, and click Export to export the JSON file (useful for the next step).
Step 2: Create MyShell Agent
After saving the workflow in Shellagen's cloud deployment, you need to create an AI robot on MyShell so that your Agent can run.
Link: https://app.myshell.ai/robot-workshop/create
Name the robot
Select ShellAgent mode
Upload JSON file
Save
At this point, 'MyShell Agent-AI Child Book' is deployed. Even if you have never used AI workflow frameworks like comfyUI before, simply following the ShellAgent tutorial can produce a finished product in two minutes, greatly lowering the threshold for building traditional agents.
Everyone can try running the 'X Agent-AI Child Book' template according to the official tutorial. This robot, X Agent-AI Child Book, can track the X account and automatically generate replies when keywords appear in the comments section.
Official tutorial for X Agent-AI Child Book starts from this section:
Create a template yourself
According to the official tutorial, you can also create a template yourself.
To create your own template, please visit: https://shellagent.myshell.ai/app
Click 'Create'
Select 'Create a Blank'
Name the agent
Choose MyShell Agent or X Agent
You will get an initial workflow, click State/Condition on the left to add 'State Flow' (working box) in the workflow, and then click the '+' on both sides to connect the 'State Flow'.
To set 'variables', please refer to paragraph nine in the white paper:
Official white paper: https://docs.myshell.ai/create/shellagent-mode/example/ai-child-book#deploy-the-x-agent
Finally, create a workflow that matches the template, then try running it; if it runs successfully, you can publish it.
Summary
Currently, besides the two official templates, 7 AI Agents have already been created using MyShell + DeepSeek. According to the official announcement, the first 50 shells will each receive 50 Shell Coins, so if you're a Conf UI workflow expert, hurry up and create. After completing your creation, please share your experiences on X and post the link in the official Discord to receive rewards.
Final Thoughts
MyShell was established in 2023 when Web2 AI was in the explosion stage of large models (LLMs), while Web3 AI was still in the rough stage of Web2 wrapping. At that time, MyShell stood out by insisting on laying out the AI Agent track, which was only validated in the second half of 2024, demonstrating the foresight of the MyShell team. After securing $6 million in angel investment in 2023, it gained another $11 million in the pre-A round at the end of March 2024, totaling $17 million.
Since MyShell follows the AI Agent app store model, the team has been deeply focused on infrastructure, causing them to miss out on the hype surrounding Web3 AI Agents represented by Virtual and ai16z.
The FOMO sentiment around the Web3 AI Agent concept has now returned to calm, indicating that after the bubble, the sector has returned to the fundamental issue of practicality.
MyShell has not rushed to issue tokens, but has consistently focused on building infrastructure, allowing AI Agent applications to mature first. By creating a no-code development framework, ShellAgent, and combining it with basic/developer/ShellAgent modes, it also provides developers with computing resources, model hosting, API calls, and other supporting services... Through these solid foundations, MyShell has accumulated 5 million users, 200,000 creators, 20,000 AI Agent applications, and a complete Agent development framework, ShellAgent, among other achievements.
By first improving the infrastructure, attracting developers to complete the ecological loop, and then using a post-issue token incentive model to maintain ecological operations, this way, when $Shell issues tokens, payments and settlements can seamlessly connect between creators, consumers, and sponsors, effectively escaping the 'mining-extraction-resale' death spiral of Web3.
Currently, the AI Agent's model of 'issuing coins first, then doing work' has been deemed unviable as the AI Agent bubble burst. MyShell's approach of 'doing work first, then issuing coins' remains to be validated. However, from an economic perspective, with mature products and a vast user base, both the demand and consumption sides have been established, and by utilizing an economic model to drive this flywheel, it may go further and more steadily.
