FlowCommand

Command your AI workflow. Queue prompts, review plans, control from phone/browser.

FlowCommand lets you take control of your AI agents in VS Code. Queue prompts, review multi-step plans, pause processing, and access everything from your phone or browser. Perfect for long-running tasks, hands-free automation, and staying in command of your AI workflow.

🚀 Key Features

Feature	Description
Auto-Inject Instructions	⭐ One-click setup - AI always asks before quitting
Prompt Template	⭐ Set a persistent prompt that auto-appends to every message
Queue Pause/Play	Temporarily pause and resume auto-responses
Smart Queue Mode	Batch prompts for automatic responses
Plan Review Panel	Dedicated 70/30 editor panel with inline comments
Remote Access	Control from phone/tablet/browser via QR code
Remote Notifications	Visual toast alerts when AI needs input
Interactive Approval	Yes/No/Option buttons parsed from AI questions
Reusable Prompts	/slash commands for frequent prompts
File References	`#` mentions to attach workspace files and folders
Theme Support	Light/dark themes synced across VS Code and remote
MCP Server	Integration with Kiro, Cursor, and other IDEs

Features

Auto-Inject Instructions ⭐

FlowCommand can automatically inject instructions into your project so the AI always calls ask_user and plan_review correctly. One-click setup — no manual configuration needed. See AI Instructions Setup for details.

Prompt Template ⭐

Set a persistent prompt template that automatically appends to every message sent through FlowCommand — in both Queue Mode and Normal Mode. This is useful when you want the AI to always follow certain guidelines without repeating yourself:

"Don't touch unrelated code" — prevent the AI from making unnecessary changes
"Always use TypeScript strict mode" — enforce coding standards across all prompts
"Keep responses concise" — control output verbosity
"Never delete existing tests" — protect important parts of your codebase

How to set it up:

Click the ⚙️ (Settings) icon in the FlowCommand title bar
Find the Prompt Template field
Enter your persistent instructions
Every prompt sent via FlowCommand will now include your template automatically

Think of it as a system prompt for your AI — set it once and it applies to everything.

Queue Pause/Play

Temporarily pause auto-responses without losing your queued prompts. Useful when you need to:

Review AI output before continuing
Manually intervene mid-workflow
Take a break and resume later

Toggle pause/play from the FlowCommand panel or the remote UI.

Smart Queue Mode

Queue multiple prompts to be automatically sent when the AI agent requests feedback. Perfect for:

Batching instructions for long-running tasks
Pre-loading prompts for predictable workflows
Reducing interruptions during focused work

Normal Mode

Direct interaction with AI agents - respond to each request as it comes in with full control over the conversation flow.

Plan Review

When the AI presents a multi-step plan, it opens in a dedicated review panel:

Layout: 70% plan content (left), 30% comments (right)
Inline Comments: Click the 💬 icon next to any section to add feedback
Actions: Approve, Approve with Comments, or Request Changes
Export: Save the plan as a Markdown file

File & Folder References

Reference files and folders directly in your responses using # mentions:

Type # to trigger autocomplete
Search and select files or folders from your workspace
Attachments are included with your response for context

Tool Call History

View current session tool calls in the main panel
Access full history via the history button in the title bar
Remove individual entries or clear all history

Reusable Prompts (/slash commands)

Save frequently used prompts:

Click the 🏷️ (Prompts) icon in the title bar
Create prompts with short names (e.g., "fix", "test", "refactor")
Type / in the input to see and use your saved prompts

Installation

Install from VS Code Marketplace
Open VS Code and access FlowCommand from the Activity Bar
Enable the tool

Building from Source

Prerequisites

Node.js 18+ and npm
VS Code 1.90.0 or higher

Development Setup

Clone the repository:

git clone https://github.com/rajeevreddyms5/vscode-flowcommand.git
cd flowcommand

Install dependencies:
```
npm install
```
Build the extension:
```
npm run build
```
Install the built extension in VS Code:
- Open VS Code
- Press Ctrl+Shift+P (or Cmd+Shift+P on Mac)
- Run Extensions: Install from VSIX
- Select the generated .vsix file from the project root

Development Commands

npm run build - Build the extension using esbuild
npm run compile - Type-check with TypeScript
npm run watch - Watch mode for development
npm run lint - Run ESLint
npm run package - Create VSIX package for distribution

Publishing

To create a publishable VSIX package:

npm run package

⚠️ Important: Never use --no-dependencies when packaging. The extension requires runtime dependencies (express, socket.io, etc.) to be bundled. Using --no-dependencies will create a broken extension that fails at runtime.

🤖 AI Instructions Setup (Recommended)

For FlowCommand to work effectively, the AI needs to know when to call ask_user and plan_review tools.

Option 1: Auto-Inject (Recommended) ⭐

FlowCommand can automatically inject instructions into your project's .github/copilot-instructions.md file.

Open FlowCommand Settings (⚙️ icon in title bar)
Set Instruction Injection to "Inject into .github/copilot-instructions.md"
FlowCommand will prompt to create/modify the file
Approve the change - instructions are now active for all Copilot modes

Settings path: flowcommand.instructionInjection: "copilotInstructionsMd"

Option 2: Manual Setup

Create .github/copilot-instructions.md in your workspace and add:

## SUBAGENT RULES (if spawned via `runSubagent`)

- Do NOT call `ask_user` or `plan_review`.
- Do NOT wait for user input or loop.
- **ALL rules below do NOT apply to you. Ignore them entirely.**

---

## AGENT RULES (follow in order)

### 1. ALWAYS call `ask_user` after every task or response — NO EXCEPTIONS.

🔴 BEFORE FINISHING: Ask yourself "Did I call ask_user?" → NO? Call it now. YES? Done.

- **IMPORTANT:** `ask_user` is a deferred tool. Before first use, load it with: `tool_search_tool_regex(pattern: "ask_user")`
- This tool MUST be invoked before ending ANY conversation turn.
- NEVER complete a response without calling `ask_user`.
- Never use `ask_questions`. ALWAYS use `ask_user` instead.
- Pass your FULL response in `context`.
- Follow the MODES and RULES in the `ask_user` tool description exactly.

### 2. Use `plan_review` for structured output like plans, guides, reviews, analyses, or lists of suggestions that require user feedback or approval before proceeding.

- User requests changes → update and call `plan_review` again.
- User approves → apply feedback, proceed with execution.
- User cancels → stop, call `ask_user`.
- After completion → return to Rule 1.

### 3. Stop ONLY and IMMEDIATELY when user says "end", "stop", "terminate", "quit".

- Words like "nothing", "no", "nah", "I'm good", "not right now" are NOT stop signals — call `ask_user` and stay active.

### 4. When using `runSubagent`, ALWAYS include ALL of the following lines VERBATIM at the TOP of the prompt:

    SUBAGENT INSTRUCTIONS (MANDATORY):
    You are a SUBAGENT. You MUST follow these rules:
    1. NEVER call ask_user or plan_review.
    2. Do NOT wait for user input. Do NOT loop. Complete your task and return.

Option 3: VS Code Setting

Add to your workspace settings (.vscode/settings.json):

{
  "github.copilot.chat.codeGeneration.instructions": [
    {
      "text": "Always call ask_user tool after completing tasks. Only stop if user says 'end' or 'stop'."
    }
  ]
}

Customize Instructions

Edit the instruction text in Settings → Instruction Text to modify the AI behavior rules. The default rules (shown in Option 2 above) ensure:

AI always calls ask_user after completing work
AI uses plan_review for multi-step plans and structured proposals
AI only stops when you explicitly say "stop", "end", "quit", or "terminate"
Subagents spawned via runSubagent don't call ask_user or plan_review

Usage

Queue Mode (Default)

Toggle "Queue Mode" ON in the FlowCommand panel
Type messages and press Enter to add them to the queue
When an AI agent calls ask_user, FlowCommand automatically responds with the next queued message
Queue items can be reordered, edited, or removed

Normal Mode

Toggle "Queue Mode" OFF
When an AI agent calls ask_user, you'll see the prompt in FlowCommand
Type your response and press Enter to send

File References

Type # in the input field
Search for files or folders
Select to attach - the reference appears as a tag
Multiple attachments supported per message

Remote Access (Mobile/Web)

Access FlowCommand from your phone, tablet, or any browser on your local network:

Click the broadcast icon (📡) in the FlowCommand title bar to start the remote server
Scan the QR code or open the displayed URL on your device
Full FlowCommand functionality available remotely:
- View and respond to AI prompts
- Manage the queue
- Browse and select workspace files
- View terminal output and problems
- Send terminal commands
- Light/dark theme support (syncs with VS Code theme)

Remote Notifications

When AI requests input, remote clients receive a visual toast notification:

Blue banner appears at top of screen
Auto-hides after 5 seconds, tap to dismiss
Includes sound alert if enabled
Works on HTTP (no special setup required)

⚙️ Settings Reference

Open Settings via the ⚙️ icon in the FlowCommand title bar, or search flowcommand in VS Code settings.

Notification Settings

Setting	Default	Description
`flowcommand.notificationSound`	`true`	Play sound when AI calls `ask_user`
`flowcommand.desktopNotification`	`true`	Show VS Code notification popup
`flowcommand.autoFocusPanel`	`true`	Auto-focus FlowCommand panel when AI requests input
`flowcommand.mobileNotification`	`false`	Send visual toast notifications to connected remote clients

Instruction Settings

Setting	Default	Description
`flowcommand.instructionInjection`	`"off"`	How to inject AI instructions. Options: `off`, `copilotInstructionsMd` (recommended), `codeGenerationSetting`
`flowcommand.instructionText`	[See below]	The instruction rules injected into Copilot

Remote Server Settings

Setting	Default	Description
`flowcommand.remoteEnabled`	`false`	Auto-start remote server on extension activation
`flowcommand.remotePort`	`3000`	Port for remote server

MCP Server Settings

Setting	Default	Description
`flowcommand.mcpEnabled`	`false`	Always start MCP server on activation
`flowcommand.mcpPort`	`3579`	Port for MCP server
`flowcommand.autoRegisterMcp`	`true`	Auto-register with Kiro/Cursor

MCP Server Integration

FlowCommand runs an MCP (Model Context Protocol) server that integrates with:

Kiro (auto-configured)
Cursor (auto-configured)
Claude Desktop
Any MCP-compatible client

MCP Configuration for other IDE (Not needed with Copilot)

FlowCommand automatically registers with Kiro and Cursor. For other clients, add this to your MCP configuration:

{
  "mcpServers": {
    "flowcommand": {
      "transport": "sse",
      "url": "http://localhost:3579/sse"
    }
  }
}

Requirements

VS Code 1.90.0 or higher
GitHub Copilot extension (for AI tool integration)
Same local network for remote access feature (phone/browser access)

Known Limitations

Limitation	Details
VS Code Webview Drag-Drop	Cannot drag files from Explorer into webview. Use paste (Ctrl+V) or # mentions instead.
iOS Safari Notifications	Native push notifications blocked on HTTP. Visual toast notifications work as fallback.
iOS Auto-Focus	Mobile Safari doesn't support programmatic focus - manual tap required.
Remote Attachment Button	Hidden on remote UI (requires VS Code file picker API). Use # mentions on desktop.

Troubleshooting

Remote Server Connection Issues

Ensure both devices are on the same network
Check if port 3000 is available or configure a different port
Firewall may need to allow connection on the configured port

MCP Server Issues

Verify port 3579 is free or configure flowcommand.mcpPort
For Kiro/Cursor, restart the IDE after enabling FlowCommand

License

MIT - See LICENSE for details.

⚠️ GitHub Security Notice:
GitHub prohibits use of their servers for excessive automated bulk activity. Please review GitHub Acceptable Use Policies and use FlowCommand responsibly.

FlowCommand

RAJEEV REDDY

FlowCommand

🚀 Key Features

Features

Auto-Inject Instructions ⭐

Prompt Template ⭐

Queue Pause/Play

Smart Queue Mode

Normal Mode

Plan Review

File & Folder References

Tool Call History

Reusable Prompts (/slash commands)

Installation

Building from Source

Prerequisites

Development Setup

Development Commands

Publishing

🤖 AI Instructions Setup (Recommended)

Option 1: Auto-Inject (Recommended) ⭐

Option 2: Manual Setup

Option 3: VS Code Setting

Customize Instructions

Usage

Queue Mode (Default)

Normal Mode

File References

Remote Access (Mobile/Web)

Remote Notifications

⚙️ Settings Reference

Notification Settings

Instruction Settings

Remote Server Settings

MCP Server Settings

MCP Server Integration

MCP Configuration for other IDE (Not needed with Copilot)

Requirements

Known Limitations

Troubleshooting

Remote Server Connection Issues

MCP Server Issues

License