Introducing TMAN: The TEN Framework Package Manager

Building real-time multimodal AI applications shouldn't require juggling multiple package managers, complex dependency chains, and fragmented tooling. That's why we created TMAN (TEN Manager) — the all-in-one command-line tool that makes developing with the TEN Framework fast, simple, and enjoyable.

Released last week, TMAN brings together project scaffolding, dependency management, visual development, and deployment into a single, powerful CLI tool. Whether you're building a voice assistant, transcription service, or complex multi-agent system, TMAN streamlines your entire workflow.

What is TMAN?

TMAN is the official package manager and development tool for the TEN Framework. Think of it as npm for Node.js, cargo for Rust, or pip for Python — but specifically designed for building multimodal AI agents.

At its core, TMAN handles:

• Package Management: Install and manage 60+ pre-built extensions (ASR, TTS, LLM, tools)
• Project Scaffolding: Create new apps and extensions from battle-tested templates
• Dependency Resolution: Automatically resolve complex dependency trees with version constraints
• Visual Development: Launch an integrated graph designer for building agent workflows
• Build & Deploy: Package and publish your extensions to the TEN registry

Getting Started in 60 Seconds

Installing TMAN is a breeze. Choose your preferred method:

macOS (via Homebrew)

brew install TEN-framework/ten-framework/tman

Linux (Ubuntu/Debian)

sudo add-apt-repository ppa:ten-framework/ten-framework
sudo apt update
sudo apt install tman

Universal Install Script

bash <(curl -fsSL https://raw.githubusercontent.com/TEN-framework/ten-framework/main/tools/tman/install_tman.sh)

The install script automatically detects your OS and architecture (x64, ARM64) and downloads the correct binary. Within seconds, you're ready to go.

Verify your installation:

tman --version

Creating Your First AI Agent

Let's build a real-time transcription app in less than 5 minutes:

# Create a new app from the transcriber template
tman create app my-transcriber --template transcriber_demo

# Navigate to your app
cd my-transcriber

# Install dependencies (extensions, systems, addons)
tman install

# Install Python/npm dependencies
tman run install_deps

# Build the application
tman run build

# Start your AI agent
tman run start

That's it! You now have a fully functional real-time transcription service running locally. TMAN handled:

• Scaffolding the project structure
• Resolving and installing extension dependencies
• Setting up the build environment
• Launching the application server

Key Features That Make TMAN Special

# Install all dependencies from manifest.json
tman install

# Install a specific extension
tman install extension deepgram_asr_python

# Install with version constraint
tman install extension openai_llm2_python@1.0.0

# Link a local extension for development
tman install ./my-custom-extension

tman designer

{
  "scripts": {
    "start": "python3 start.py",
    "build": "npm run build && python3 build_extensions.py",
    "test": "pytest tests/",
    "deploy": "docker build -t my-agent . && docker push my-agent"
  }
}

tman run start
tman run build
tman run test -- --verbose  # Pass extra args after --

# Package your extension
tman package
# Creates .ten/package/my_extension.tpkg

# Publish to the TEN registry
tman publish

Real-World Use Cases

Voice Assistant Development

# Create a voice assistant with turn detection
tman create app voice-agent --template voice-assistant-with-turn-detection

cd voice-agent
tman install

# Launch the designer to customize the pipeline
tman designer

# Run the agent
tman run start

Extension Development Workflow

# Create a custom extension
tman create extension my_sentiment_analyzer --template default_extension_python

cd my_sentiment_analyzer

# Edit your extension code
# ... implement your logic ...

# Test locally by linking to an app
cd ../my-app
tman install ../my_sentiment_analyzer

# Package and publish when ready
cd ../my_sentiment_analyzer
tman package
tman publish

CI/CD Integration

TMAN works great in automated pipelines:

FROM ubuntu:22.04

# Install tman
RUN curl -fsSL https://github.com/TEN-framework/ten-framework/releases/download/0.11.42/tman-linux-release-x64.zip -o tman.zip \
    && unzip tman.zip \
    && cp ten_manager/bin/tman /usr/local/bin/

# Install dependencies
COPY manifest.json .
RUN tman install --production

# Build and run
COPY . .
RUN tman run build
CMD ["tman", "run", "start"]

Under the Hood

TMAN is built with Rust, giving it exceptional performance and reliability. Here's what powers it:

• clap: For intuitive command-line parsing
• actix-web: Serves the designer web interface
• tokio: Asynchronous runtime for concurrent operations
• semver: Handles semantic versioning logic
• reqwest: Communicates with the TEN package registry

The result? A blazingly fast CLI that handles complex operations in milliseconds.

Platform Support

TMAN runs everywhere you need it:

• Linux x64 (Ubuntu, Debian, Fedora, etc.)
• Linux ARM64 (Raspberry Pi, ARM servers)
• macOS Intel (x64)
• macOS Apple Silicon (ARM64)
• Windows x64

Pre-built binaries are available for all platforms on our GitHub Releases.

Command Reference

Here's a quick reference of the most commonly used commands:

For detailed usage, run:

tman --help
tman <command> --help

Shell Completion

TMAN supports shell completions for a better CLI experience:

# Bash
tman completion bash > ~/.local/share/bash-completion/completions/tman

# Zsh
tman completion zsh > /usr/local/share/zsh/site-functions/_tman

# Fish
tman completion fish > ~/.config/fish/completions/tman.fish

# PowerShell
tman completion powershell | Out-String | Invoke-Expression

What's Next?

TMAN is just getting started. Here's what we're working on:

• Extension templates marketplace: Browse and use community templates
• Improved conflict resolution: Better error messages when dependencies conflict
• Plugin system: Extend TMAN with custom commands
• Performance optimizations: Even faster dependency resolution
• Enhanced designer: More visual features and integrations

Get Involved

TMAN is open-source and we'd love your contributions! Here's how to get started:

1. Try it out: Install TMAN and build your first AI agent
2. Share feedback: Open issues on GitHub
3. Contribute: Submit PRs for features or bug fixes
4. Build extensions: Create and publish extensions to grow the ecosystem

Conclusion

TMAN brings professional-grade package management to the TEN Framework, making it easier than ever to build real-time multimodal AI applications. Whether you're prototyping a voice assistant or deploying a production-grade multi-agent system, TMAN handles the complexity so you can focus on building amazing experiences.

Ready to get started? Install TMAN today:

# macOS
brew install TEN-framework/ten-framework/tman

# Linux
sudo add-apt-repository ppa:ten-framework/ten-framework
sudo apt install tman

# Universal
bash <(curl -fsSL https://raw.githubusercontent.com/TEN-framework/ten-framework/main/tools/tman/install_tman.sh)

Join our community:

• GitHub
• Documentation
• Discord

Happy building! 🚀