AI+

Orchestrate AI coding agents with any OpenAI-compatible LLM — plan, delegate, validate, and ship.

AI+ is a VS Code extension that turns your LLM into a structured coding orchestrator. It connects to any OpenAI-compatible API endpoint, analyzes your project, breaks complex tasks into small validated subtasks, and delegates implementation to your configured coding agent — all within VS Code.

Features

Orchestrator + Agent Architecture

AI+ never edits files directly. It acts as a project manager:

Researches your codebase with read-only tools (grep, file reading, search)
Splits big tasks into small, independent subtasks
Delegates one subtask at a time to your configured coding agent
Validates each subtask's output before moving to the next
Reports the final result with a summary of all changes

Inline AI! Completion

Write a comment with an instruction and press Shift+Tab:

// implement a retry handler with exponential backoff AI!

AI+ processes the instruction inline, applies the edit, verifies it for syntax errors, fixes any issues, and plays a success animation when done. If the primary model fails, it automatically falls back to your configured agent.

Project Setup (`/setup`)

Run /setup in the AI+ chat to map your workspace:

Skill files under .agents/skills/ — one per relevant technology (unit tests, YAML, Python, API patterns, etc.)
projectstructure.md — directory layout and what lives where
agent.md — project summary with links to every skill file

Bring Your Own Everything

Any OpenAI-compatible model — no bundled provider lock-in
Any coding agent — configure via a single setting
Custom system prompts — add project or org-specific guidance
Configurable timeouts — separate plan and execute timeout settings

Neumorphism UI

Modern soft-shadow design system applied to all panels, buttons, inputs, and dialogs — works in both light and dark themes.

Sound Notifications

Optional chime when a task completes so you get an audible cue to review results.

Getting Started

1. Install

Install the extension from the VS Code Marketplace.

2. Configure Your LLM

Open AI+ from the Activity Bar and enter:

Base URL — your OpenAI-compatible API endpoint
API Key — your authentication key
Model Name — the model to use

Supported providers include any endpoint that speaks the OpenAI chat-completions shape: hosted providers, local gateways, enterprise proxies, Ollama, LM Studio, and more.

3. Configure Your Coding Agent

In Settings → AI+, set the Subagent Command to the CLI you use for code implementation. This is the tool that AI+ delegates all file changes to.

4. Run `/setup`

Open a workspace and type /setup in the AI+ chat. AI+ will analyze your project and generate skill files and a project map.

5. Start Building

Ask AI+ to implement features, fix bugs, write tests, or refactor code. It will split the work into subtasks, delegate each one, and validate the results.

Usage

Chat-Based Tasks

Type any request in the AI+ chat panel:

Request Type	Example	What Happens
Implementation	"Add a user registration API"	AI+ splits into model → repository → service → controller → tests, delegates each to your agent, validates, and reports
Research	"Explain how the auth system works"	AI+ handles alone with read-only tools — no agent calls
Bug fix	"Fix the null pointer in UserService"	AI+ delegates to your agent with a precise task spec
Refactoring	"Extract the database logic into a separate module"	AI+ plans the split, delegates each piece, validates

Inline AI! Completion

Write a comment with your instruction followed by AI!:
```
# add input validation for email field AI!
```
Press Shift+Tab
AI+ applies the edit, verifies it, and shows a green success animation
The AI! comment is automatically removed after a successful edit

Validation Loop

After every delegated subtask, AI+ independently verifies the work:

Reads the changed files
Checks against the acceptance criteria
If validation fails → re-delegates with a corrective prompt (up to 3 retries)
Only reports success after independent verification

Settings

Setting	Default	Description
`aiplus.subagentCommand`	`pi`	The coding-agent CLI that AI+ delegates implementation work to
`aiplus.customSystemPrompt`	(empty)	Extra instructions appended to AI+'s built-in orchestrator prompt
`aiplus.playSoundOnTaskComplete`	`true`	Play a chime when a task completes
`aiplus.idfcCoderPlanTimeout`	`20`	Timeout in minutes for plan-mode delegation
`aiplus.idfcCoderExecuteTimeout`	`30`	Timeout in minutes for execute-mode delegation

Keyboard Shortcuts

Shortcut	Command	Description
`Shift+Tab`	`AI+: Implement AI! Instruction at Cursor`	Process an inline `AI!` comment at the cursor position

Requirements

VS Code 1.101.0 or later
An OpenAI-compatible LLM endpoint (API key + base URL + model name)
A coding-agent CLI installed and available on your PATH (for implementation delegation)

Extension Settings

All settings are found under Settings → Extensions → AI+:

Subagent Command — The CLI command AI+ uses to delegate implementation work
Custom System Prompt — Optional extra instructions for project/org-specific guidance
Play Sound on Task Complete — Audible notification when tasks finish
Plan Timeout — Minutes to wait for plan-mode delegation (default: 20)
Execute Timeout — Minutes to wait for execute-mode delegation (default: 30)

Known Issues

Inline AI! completion may not work in read-only editors
Very large files (>10k lines) may slow down the context gathering for inline completion

Release Notes

2.0.0

Orchestrator + Agent architecture — AI+ now acts as a project manager that splits big tasks into small subtasks, delegates one at a time, and validates each before proceeding
Inline AI! completion — Write # instruction AI! and press Shift+Tab for inline code generation with post-edit verification and success animation
Custom pi tools — Two dedicated LLM tools (pi_plan and pi_execute) with prompt sanitization and configurable timeouts
Strict task specification format — Delegated tasks follow a 12-section structured format for reliable implementation
Neumorphism UI — Modern soft-shadow design system across all components
Sound notifications — Optional chime on task completion
Custom system prompt — User-configurable additional instructions
Post-edit verification — Reads file after edit, checks syntax errors, fixes stray artifacts, removes trigger comment
Success/failure animations — Green glow on successful edit, red flash on failure

1.0.0

Initial release of AI+
OpenAI-compatible LLM endpoint support
Project /setup with skill files and project structure mapping
Chat-based task delegation with plan + execute workflow
Bring-your-own-agent architecture

License

Apache-2.0

Repository

GitHub

AI+

simpletools.in

AI+

Features

Orchestrator + Agent Architecture

Inline AI! Completion

Project Setup (/setup)

Bring Your Own Everything

Neumorphism UI

Sound Notifications

Getting Started

1. Install

2. Configure Your LLM

3. Configure Your Coding Agent

4. Run /setup

5. Start Building

Usage

Chat-Based Tasks

Inline AI! Completion

Validation Loop

Settings

Keyboard Shortcuts

Requirements

Extension Settings

Known Issues

Release Notes

2.0.0

1.0.0

License

Repository

Project Setup (`/setup`)

4. Run `/setup`