Verdent & VS Code Extension: This IS SO GOOD! This AI Coding AGENT combines GPT-5 & Claude!

Key Concepts

• Verdant: An AI coding agent designed as a professional coding assistant for large codebases.
• Production-ready code: Code that is thoroughly reviewed, bug-free, and ready for deployment.
• Desktop App: The most feature-rich version of Verdant, supporting multiple agents and advanced functionalities.
• VS Code Extension: A lighter version of Verdant, primarily for pair programming, integrated directly into the VS Code editor.
• Credits: The currency used to operate Verdant, with different pricing plans offering varying amounts.
• Reasoning Levels: Options (none, standard, high, maximum) to control the depth of analysis and planning by the AI.
• Plan Mode: A feature that allows for multi-turn planning conversations to define tasks before execution.
• Models: Different AI models (Sonnet, Haiku, GPT-5, Claude, Gemini, MiniMax M2) that can be selected for various tasks.
• Work Tree: An option that creates a new Git branch for each task, facilitating easier review and rollback.
• Project Folder: An option that applies edits directly to the project files without branching.
• Diff Lens: An AI-powered feature that analyzes code changes (diffs) and explains what has been modified.
• Code Review: An AI-generated assessment of code quality, security, and potential fixes.
• Agentic System: The underlying architecture that allows Verdant to manage and orchestrate multiple AI agents.
• Sub-agents: Smaller, specialized agents that can be spawned by the main agent with custom contexts.
• Browser Use Tool: A feature in the VS Code extension that allows the AI to test applications it generates.

Verdant AI Coding Agent: A Comprehensive Overview

This video provides an in-depth look at Verdant, a new AI coding agent designed to function as a professional coding assistant, particularly for large codebases. The presenter highlights Verdant's capability to generate production-ready code from simple prompts, ensuring bug-free and reviewed output.

Verdant's Offerings and Pricing

Verdant is available in two main forms:

• Desktop App: This is the most comprehensive version, offering a full suite of features, including the ability to spin up multiple agents simultaneously.
• VS Code Extension: This version is more focused on pair programming and integrates directly into the VS Code editor. While it doesn't support multiple agents in the same way as the desktop app, it allows for shared subscriptions and agentic capabilities.

The pricing structure includes three tiers:

• $20 Plan: Offers a significant amount of credits.
• $60 Plan: Provides a larger credit allocation.
• $180 Plan: The highest tier, offering the most credits.

Currently, Verdant is offering double the credits for each plan. Additionally, a 7-day free trial is available, granting 200 credits for use with any model, allowing users to test the platform before committing.

Desktop App Functionality and Workflow

The desktop app's interface is designed for clarity and efficiency.

• Main Screen: Features a prominent prompt box where users can input tasks. It also allows for project selection and the addition of image context for design purposes.
• Reasoning Levels: Users can select from "none," "standard," "high," and "maximum" reasoning levels to tailor the AI's analytical depth.
• Plan Mode: This feature enables a multi-turn planning conversation, where the AI asks follow-up questions to refine the implementation strategy before code generation begins.
• Model Selection: Users can choose between various AI models, including Sonnet, Haiku, and GPT-5, for different tasks.
• Work Tree vs. Project Folder:
  • Work Tree: Creates a new Git branch for each task, simplifying the review of changes and enabling quick rollbacks if issues arise. This is ideal for maintaining a clean Git history.
  • Project Folder: Edits are applied directly to the project files without creating branches. This offers a more direct modification approach.
• Task Management: The bottom section of the interface displays and manages tasks, especially useful when multiple agents are active.

Practical Demonstration: FPS Game Enhancement

The presenter demonstrates Verdant's capabilities by tasking it with enhancing a Godot-based FPS game. The specific requests were to:

1. Add a step counter.
2. Implement a life bar affected by jumping.

Verdant successfully completed these tasks in a single attempt, showcasing its speed and integration with models like GPT-5 and Sonnet.

Key Features of the Desktop App Workflow:

• Status Chip and Sidebar: A "status chip" opens a sidebar that details exactly which files are being edited, keeping the main UI uncluttered.
• Manual File Editing: Users can navigate and manually edit files within the left-hand file explorer.
• External Editor Integration: The option to open projects in external editors is available.
• Git Repo Setup: The tool can set up a Git repository if one doesn't exist.
• Integrated Terminal: A terminal is accessible directly within the app for command-line operations.
• Diff Lens: A unique feature that provides an AI-generated analysis of code changes, highlighting modified lines and explaining their purpose. This eliminates the need for manual git diff checks.
• Code Review Generation: With each change, users can generate an AI-powered code review using the GPT-5 model. This review assesses security, code quality, and suggests fixes.
• Direct Committing: Changes can be committed directly from the interface, keeping the Git repository up-to-date.
• Tabbed Project Management: Projects are managed in tabs, similar to a web browser, allowing for easy navigation between different repositories and their associated tasks.
• Agent Markdown File: Verdant adheres to a markdown file within the agent's directory, allowing for project context and rule definitions.
• Build Script Configuration: Users can specify build scripts and commands in the settings so Verdant knows how to run applications correctly.
• Global Settings: Includes options for proxy setup, global rules, permissions, and a deny list for commands. Users can also configure default behaviors like automatic code reviews or commits, and switch code review models (GPT-5 or Gemini). Display preferences like font size are also adjustable.

VS Code Extension Features

The VS Code extension offers a streamlined experience for developers:

• Installation: Easily installable via the VS Code marketplace.
• Pair Programming Focus: Primarily designed as a pair programming tool.
• Agentic Capabilities: While not supporting multiple agents simultaneously like the desktop app, it leverages the same agentic capabilities and allows for subscription sharing.
• Rule Support: Supports defining rules for the AI's behavior.
• Sub-agents with Custom Context: The main agent can spawn sub-agents with specific contexts, enabling more complex task orchestration.
• Browser Use Tool: A notable feature that allows the AI to test the applications it generates directly within a browser.
• Model Configuration: Users can select models like Claude, GPT-5, and the recently added MiniMax M2 (free until November 7th). MiniMax M2 is also slated for inclusion in the desktop app.

Presenter's Thoughts and Conclusion

The presenter is highly impressed with Verdant's performance, citing it as one of the best-performing AI agents on their benchmarks. They highlight its ability to handle complex tasks, such as generating a movie tracker app in one go.

Key takeaways from the presenter's perspective:

• Ideal for Developers: Verdant is particularly beneficial for developers who use pair programming tools like Claude Code and are concerned about code base management or the complexity of using multiple agents.
• Efficient Agent Orchestration: The system excels at spinning up and managing sub-agents, providing clear code reviews, and ensuring understanding of generated code.
• Professional and Memory Efficient: Verdant is described as a professional system that is also memory efficient, offering superior capabilities to tools like Claude Code without being a resource hog.
• Unified Subscription: The availability of both a multi-agent desktop app and a daily-use VS Code extension under a single subscription provides excellent value.
• Thoughtful Design: The tool demonstrates a high level of thoughtful design and implementation.

In conclusion, Verdant is presented as a powerful, professional, and user-friendly AI coding agent that significantly enhances developer productivity and code quality.