First Impressions of Verdent: An AI Coding Assistant Supporting Subagents

Verdent is a VS Code plugin positioned as a “senior developer assistant for the AI era,” featuring Subagent concurrent tasks and automatic code quality validation. This article shares my hands-on experience, analyzes its strengths and weaknesses, and discusses its practical value in developer workflows.

Recently, I received a trial qualification for Verdent and installed the plugin in VS Code. Verdent is the first product launched by Codeck, aiming to help developers grow more efficiently in the AI era. The official slogan is: “Human developers will thrive in the AI era,” emphasizing the importance of collaboration between human programmers and AI. Verdent was recently released, and you can learn more on the official website: https://www.verdent.ai/ .

There are many Vibe Coding tools and IDE plugins on the market, and I have tried quite a few. Initially, I didn’t expect much from Verdent, but after several days of use, I found it does have some highlights and areas for improvement worth sharing.

What I Like

Using Verdent as a VS Code plugin, several aspects impressed me:

• Natural integration: Install the plugin and start using it—no need to learn a new workflow, lowering the barrier to entry.
• Beyond code completion: Not just code completion; it can automatically break down tasks, generate solutions, and validate code quality.
• Strong context awareness: Able to respond based on project code, not limited to the current cursor position.
• Subagent support: Verdent’s biggest highlight is supporting concurrent tasks for a single agent, with configurable subagents and FailFast code checks for validation.
• Fast response: Compared to tools like Codex, Verdent generates results faster, improving development efficiency.

Areas That Need Improvement

Although Verdent is powerful, there are some details in the user experience that could be better:

• Cumbersome context selection: Unlike Copilot, which allows quick selection with simple symbols, Verdent’s @ selection is powerful but could benefit from one-click selection of open files or terminal output. Keyboard selection of prompt history and custom prompts is supported, but the overall experience could be smoother—this is a common issue among AI coding plugins.
• Unnatural input box interaction: You can enter content but can’t send it directly, making the operation less convenient. If it worked like a regular chat, allowing instant input and sending, the experience would be smoother.
• Inconvenient command execution: Verdent automatically runs some commands, but results are shown in a panel rather than directly in the terminal. If you need to copy commands to run them, the process is a bit cumbersome.
• No Tab code completion: Interaction only happens in the panel; it can’t be used as a traditional code completion tool.
• Lack of model options: Like Qoder, there’s no model selection feature, so users don’t know which model is running—transparency is lacking.

Overall, Verdent’s core capabilities are solid, but there’s room for optimization in user experience details.

Problems They Aim to Solve

According to official materials, most AI coding tools today generate code quickly but with unstable quality, requiring manual debugging. Verdent tries to improve this by:

• Using multi-turn conversations to clarify requirements when they’re vague.
• Automatically breaking down tasks and designing solutions.
• Self-checking code after generation, generating tests, and striving to deliver runnable results.

In other words, Verdent focuses more on “delivering usable code” rather than just “generating a few lines of code.”

Two Product Forms

Verdent currently offers two product forms for different development scenarios:

• Verdent for VS Code: The plugin version I tested, suitable for developers working close to code. Low learning curve and quick to get started.
• Verdent Deck: Desktop application version, supports running multiple agents in parallel, with task boards and DiffLens features. Suitable for developers managing multiple projects simultaneously. Currently only supports M-series Macs; Windows version expected in October.

Summary

Verdent still has room for improvement in user experience details, but its overall positioning is closer to a “developer assistant” rather than a traditional “code completion tool.” Features like Subagent concurrent tasks and automatic code quality validation offer developers a more efficient collaboration experience.

If you’re interested in AI coding assistants, Verdent is worth keeping an eye on—there may be more innovation and feature improvements in the future.