Coplay integrates seamlessly with Unity through an intuitive interface designed to enhance your game development workflow. This guide explains all the UI elements you’ll encounter when using Coplay.

​
Context

The Context dropdown determines what information Coplay has access to when responding to your queries:
  • File Tree: Provides access to your project’s file structure
  • Project Settings: Gives context about your Unity project configuration
  • Active Scene: Includes information about the currently open scene
  • Git History: Provides access to version control information
  • Packages: Shows information about installed packages
Example usage: Select “Active Scene” when asking Coplay to help optimize game objects in your current scene, or select “Project Settings” when troubleshooting configuration issues.

​
Mode

Coplay offers several modes that control how tasks are processed. Modes change often, so check the Discord release notes for up‑to‑date best practices. The available modes are:
  • Normal (default) – Balanced performance that automatically executes a sequence of actions. Suitable for most tasks.
  • StepByStep – Pauses after every action and waits for your feedback before continuing. Use this for tasks that require precise placement or manual oversight; slower but more accurate.
  • OneShotSceneGen – Generates an entire scene in a single prompt. Primarily used for large scene generation or when providing your own 3D models. This mode is being deprecated since Normal mode now handles most scene generation tasks.
  • Experimental – Includes cutting‑edge features such as batch tool calls that speed up execution. This mode is intended for testing and may be less stable.
  • PipelineRecording – Records every action in a session so you can replay the sequence or generate variations later. After recording, ask Coplay to “create a variation” to automatically apply changes based on the recorded pipeline.
  • Orchestrator – Reads tasks from a CoplayPlan.md file and executes them in order. This autopilot mode aims to transform a design document into a working game. It is powerful but still experimental and may take longer to complete tasks.
When in doubt, start with Normal mode and adjust based on your use case.

​
Model

Coplay lets you choose between the latest AI models from major foundational providers like OpenAI, Anthropic, Google, and X.AI. Each model offers distinct advantages for different Unity development tasks:

​
Anthropic

  • claude-3-7-sonnet: (Default) Balanced model with exceptional reasoning capabilities; excellent for complex architectural decisions, debugging logic errors, and explaining game mechanics with detailed, well-structured responses. More thorough but slower, taking a step-by-step approach to problem solving.
  • claude-3-7-thinking: Advanced reasoning model with enhanced step-by-step thinking capabilities, ideal for complex problem-solving tasks and detailed analysis of Unity systems. Excels at iterative thought processes.
  • claude-4-sonnet: Newer general model with improved performance and understanding; offering superior code generation and architectural guidance for Unity projects. Excellent for coding tasks but slower in tool execution.
  • claude-4-thinking: Advanced reasoning variant of Claude-4 with enhanced analytical capabilities, perfect for complex debugging and system design challenges that require thorough, methodical analysis.

​
Inception

  • mercury-coder-small: Experimental model that uses diffusion-based inference to generate code incrementally, resulting in significantly faster execution of sequential programming tasks.
  • mercury-coder-small-stream: Streaming variant that leverages diffusion techniques to provide real-time response capabilities with enhanced speed benefits.

​
OpenAI

  • gpt-4.1: Handles tasks in bulk and excels at UI development, image processing, and multiple tool execution. Can work quickly but may perform less well with complex coding tasks. Best for UI-focused work and batch operations.
  • o3: Fast, cost-effective model for straightforward development tasks, good for quick syntax questions, simple script generation, and basic Unity workflow assistance during prototyping phases.
  • o4-mini: Optimized for quick responses with good balance of speed and capability, effective for common Unity tasks like component setup, basic animations, and UI scripting with concise, practical answers.

​
X.AI

  • grok-3: Excels at creative problem-solving and unconventional solutions.

​
Google

  • gemini-2-5-pro: Strong multimodal capabilities that excels at analyzing visual elements, scene composition, and asset organization. Good with well-structured, multi-step requests and can handle actions in bulk. Offers a balanced approach between Claude’s thoroughness and GPT-4.1’s speed.
  • gemini-2-5-flash: Optimized for speed and efficiency with good baseline understanding of Unity concepts, ideal for quick reference questions, syntax checking, and basic troubleshooting when immediate guidance is needed.

​
Action Recording

Coplay can record your actions in the Unity Editor, allowing you to:
  1. Track movements and changes to game objects
  2. Delete specific recorded actions using the ”×” button
  3. Clear all recorded actions at once
  4. Save recordings for future reference
When you’ve finished recording actions, you can save them with a custom name and use them for additional context in future tasks.
Example usage: Record a sequence of object placements to document your level design process, then ask Coplay to analyze the pattern or suggest improvements.

​
Memory (configurability coming soon…)

Long-term memory is currently stored server side, we will soon release support for listing and continuing prior conversations