Ship features faster with a structured AI-assisted workflowEvymind is a structured AI workflow that lets you grow from simple prompts to refined roles and apply them as reviewable code changes.Organize work as projects and task treesAttach exactly the files you want the AI to seeReview patch proposals before applying them to your codebaseAll prompts, roles, history and results are saved as json files, so your setup survives editor restarts and can be reused across features.You connect your own OpenAI API key. There is no extra subscription: the editor talks directly to the API using your key, so you stay in control of models, policies and costs.With Evymind, you can:Ship features faster with prompt and patches from inside the editorReuse your best prompts and roles instead of rewriting them every timeImprove your context and role description to avoid recurrent LLM mistakesImprove LLM accuracy by sending precise file context easilyDesign multi-step chains for complex tasksKeep every AI change saved, visible, reviewable and under source controlThis package focuses on code and workflow, not magic editing.It is designed to help you think, design, implement and iterate code faster, while keeping you in charge of every change.This package IS:Editor-only tools for Unity developersA workflow layer around your own OpenAI accountBuilt to save prompts, roles, chains and tasks per projectThis package IS NOT:A one-click GameObject or scene auto-editorA visual content generator (no textures, models, audio)A replacement for source control or code reviewIf you want a new way of developing — simple prompts when you need speed or customizable automatic chains when you need depth.Evymind turns your editor into an AI-powered workflow that lets you design freely the way you work with LLM.Media note: The promotional video uses motion graphics and is not captured from the Unity Editor.Key featuresStructured AI workflow integrated in the editorThree editor windows: Task Graph, Request Input, Request ResultRoles & chain roles as ScriptableObjects for reusable AI behaviorsPrompt results as reviewable code patches with colored diffs before applyPer-project state files for tasks, prompts, roles, history and resultsToken usage estimate preview per request and per input fileConsole Autofix: generate “fix this error” prompts from compile/build logsParallel execution of multiple AI requestsUses your own OpenAI API key – no extra subscription, no external serverOptionaly implement your custom API key retrieval method for CI/CD or security improvements.WorkflowPlan work as projects and nested tasks in the Task Graph windowFor each task, write a prompt, pick a role or chain step, select files and history depthSend the request and monitor its status directly in the editorReview replies and proposed patches in the Request Result window, then apply or discardRoles & promptsCreate roles like Gameplay Developer, Online Specialist, Debugger, Code Reviewer, Tech Lead, etc.Store roles as assets you can version and share with your teamOptionally define chain roles with ordered steps for more structured workflowsIterate on prompts and roles to reduce LLM mistakes and standardize your coding styleFiles & patchesSelect files and folders easily through a dedicated file selection windowSee per-file token estimates before sending context to the LLMParse AI output into explicit file operations (create / modify / delete / patch)Validate and preview patches before applying; failed patches are clearly reportedWorks alongside your usual Git / version control with no hidden modificationsCompatibility & requirementsUnity: 2022.3 LTS or newerScope: editor-only; no runtime components added to your buildsPlatforms: tested in the Unity editor on Windows and macOSRequires an OpenAI account and API key (billed directly by OpenAI)Newtonsoft.Json package as a dependency (com.unity.nuget.newtonsoft-json) installed via Package Manager
