DevLog: Godot AI Suite 2.0 - From Co-Pilot to Agent

DevLog: Godot AI Suite 2.0 - From Co-Pilot to Agent

Hey everyone,

Today marks the biggest evolution for the Godot AI Suite since its inception. Version 2.0 is here, and it fundamentally changes how you can interact with AI in your development workflow. This update is the culmination of a roadmap that started with a simple question: "What if we could close the loop?"

The Journey from 1.x: The "Consultation" Era

In versions up to 1.3, the Godot AI Suite was designed as the ultimate AI co-pilot. The workflow was powerful but distinctly manual:

1. Generate: Create a Masterprompt.txt with your full project context.
2. Consult: Have an intelligent, context-aware conversation with an AI like Gemini, ChatGPT or Claude.
3. Implement: Manually copy the code, create the files, and apply the suggestions back in the Godot editor.

This was a massive time-saver. The Prompt Library in v1.3 further streamlined the "Consult" step. But we knew there was still a major point of friction: the manual implementation. Every copy-paste, every new script creation, every line change was a potential source of human error and a break in creative flow.

Our goal was to bridge that final gap.

Introducing Version 2.0: The "Execution" Era with Agent Mode

The roadmap from v1.3 to v2.0 was focused on a single, transformative feature: turning the AI from a passive consultant into an active agent.

This required a complete rethinking of the plugin's architecture and the AI's role. Here’s what we've built:

• [Feature] Dual Masterprompt System:
  The core of the new system. When you click "Generate Masterprompt," you now get two files:
  • Masterprompt_talk.txt: The classic prompt for when you want to brainstorm, ask questions, and get advice.
  • Masterprompt_agent.txt: A new, specialized prompt that instructs the AI to respond not with conversational text, but with a structured, machine-readable JSON "execution plan."
• [Feature] The AI Agent Window:
  This is the new heart of the plugin. Instead of just copying code, you now copy the AI's entire response and paste it into the Agent. The Agent parses the execution plan and presents you with a clear, step-by-step list of proposed changes.
• [Feature] Interactive Review & Execution:
  You are always in control. The Agent allows you to:
  • See the full plan: Review every file creation, deletion, or modification before it happens.
  • View Code Diffs: For any code change, a built-in diff viewer shows you exactly what lines will be added or removed, with syntax highlighting.
  • Accept with Confidence: Apply changes one-by-one or accept all automated tasks at once. The Agent handles the file operations, letting you stay focused on development.
• [Refactor] A New, Dockable UI:
  To accommodate this new workflow, we've moved the plugin's controls from a simple menu bar to a proper dockable toolbar on the right. This gives you instant access to the Settings, Prompt Library, and the new Agent window.

The New Workflow: Your Choice

With version 2.0, you now have two distinct, powerful workflows:

• Talk Mode (Consultation): Use Masterprompt_talk.txt for high-level architectural questions, brainstorming mechanics, or getting creative ideas.
• Agent Mode (Execution): Use Masterprompt_agent.txt for concrete tasks. "Refactor player.gd into a state machine." "Create a resource script for enemy stats." "Add a health bar to the main scene."

This update represents our vision for the future of AI-assisted development. A seamless cycle of high-level command, intelligent planning, and safe, automated execution.

Thank you for your continued support. We can't wait to see what you build with the new Agent!