tobi/try-cli

try - fresh directories for every vibe

Your experiments deserve a home. 🏠

C implementation of try - blazing fast, zero dependencies, same philosophy

Prefer Ruby? Use the original Ruby version - it's excellent!

Ever find yourself with 50 directories named test, test2, new-test, actually-working-test, scattered across your filesystem? Or worse, just coding in /tmp and losing everything?

try is here for your beautifully chaotic mind.

What it does

Instantly navigate through all your experiment directories with:

• Fuzzy search that just works
• Smart sorting - recently used stuff bubbles to the top
• Auto-dating - creates directories like 2025-11-30-redis-experiment
• Native speed - C binary, starts in milliseconds
• Zero dependencies - single statically-linked binary

Quick Start

Pre-built binaries (fastest)

# Linux x86_64
curl -sL https://github.com/tobi/try-cli/releases/latest/download/try-linux-x86_64.tar.gz | tar xz
sudo mv try /usr/local/bin/

# macOS (Intel)
curl -sL https://github.com/tobi/try-cli/releases/latest/download/try-darwin-x86_64.tar.gz | tar xz
sudo mv try /usr/local/bin/

# macOS (Apple Silicon)
curl -sL https://github.com/tobi/try-cli/releases/latest/download/try-darwin-aarch64.tar.gz | tar xz
sudo mv try /usr/local/bin/

Shell integration

# Add to your shell (bash/zsh)
echo 'eval "$(try init ~/src/tries)"' >> ~/.zshrc

# For fish shell users
echo 'eval (try init ~/src/tries | string collect)' >> ~/.config/fish/config.fish

Build from source

git clone https://github.com/tobi/try-cli.git
cd try-cli
make
sudo make install

# Then add shell integration (see above)

The Problem

You're learning Redis. You create /tmp/redis-test. Then ~/Desktop/redis-actually. Then ~/projects/testing-redis-again. Three weeks later you can't find that brilliant connection pooling solution you wrote at 2am.

The Solution

All your experiments in one place, with instant fuzzy search:

$ try pool
→ 2025-11-28-redis-connection-pool    2h ago, 18.5
  2025-11-03-thread-pool              3d ago, 12.1
  2025-10-22-db-pooling               2w ago, 8.3
  + Create new: pool

Type, arrow down, enter. You're there.

Features

🎯 Smart Fuzzy Search

Not just substring matching - it's smart:

• rds matches redis-server
• connpool matches connection-pool
• Recent stuff scores higher
• Shorter names win on equal matches

⏰ Time-Aware

• Shows how long ago you touched each project
• Recently accessed directories float to the top
• Perfect for "what was I working on yesterday?"

🎨 Pretty TUI

• Clean, minimal interface
• Highlights matches as you type
• Shows scores so you know why things are ranked
• Dark mode by default (because obviously)

📁 Organized Chaos

• Everything lives in ~/src/tries (configurable via --path or TRY_PATH)
• Auto-prefixes with dates: 2025-11-30-your-idea
• Skip the date prompt if you already typed a name

Shell Integration

• Bash/Zsh:

  # Default is ~/src/tries
  eval "$(try init)"
  # Or pick a path
  eval "$(try init ~/code/experiments)"
  

• Fish:

  eval (try init | string collect)
  # Or pick a path
  eval (try init ~/code/experiments | string collect)
  

Notes:

• The runtime commands printed by try are shell-neutral (absolute paths, quoted). Only the small wrapper function differs per shell.

Usage

try                                          # Browse all experiments
try redis                                    # Jump to redis experiment or create new
try clone https://github.com/user/repo.git  # Clone repo into date-prefixed directory
try https://github.com/user/repo.git        # Shorthand for clone (same as above)
try --help                                   # See all options

Git Repository Cloning

try can automatically clone git repositories into properly named experiment directories:

# Clone with auto-generated directory name
try clone https://github.com/tobi/try.git
# Creates: 2025-11-30-try

# Clone with custom name
try clone https://github.com/tobi/try.git my-fork
# Creates: 2025-11-30-my-fork

# Shorthand syntax (no need to type 'clone')
try https://github.com/tobi/try.git
# Creates: 2025-11-30-try

Supported git URI formats:

• https://github.com/user/repo.git (HTTPS GitHub)
• [email protected]:user/repo.git (SSH GitHub)
• https://gitlab.com/user/repo.git (GitLab)
• [email protected]:user/repo.git (SSH other hosts)

The .git suffix is automatically removed from URLs when generating directory names.

Keyboard Shortcuts

• ↑/↓ - Navigate
• Enter - Select or create
• Backspace - Delete character
• ESC - Cancel
• Just type to filter

Configuration

Set TRY_PATH to change where experiments are stored:

export TRY_PATH=~/code/sketches

Or use the --path flag:

eval "$(try init ~/code/sketches)"

Default: ~/src/tries

Arch Linux

Install from the AUR using your preferred helper:

yay -S try-cli
# or
paru -S try-cli

Then add shell integration to your config (see Shell Integration above).

Nix

Quick start

nix run github:tobi/try-cli
nix run github:tobi/try-cli -- --help
nix run github:tobi/try-cli -- init ~/my-tries

Nix Flakes

nix profile install github:tobi/try-cli

Home Manager

{
  inputs.try-cli.url = "github:tobi/try-cli";

  # Add to your home.nix or wherever you configure packages
  home.packages = [ inputs.try-cli.packages.${system}.default ];

  # Shell integration will be handled by your shell config
}

Why C?

The Ruby version is excellent and perfectly usable. This C port exists for:

• Speed - Instant startup, even on slow hardware
• Portability - Single binary, no runtime dependencies
• Systems without Ruby - Works anywhere there's a C compiler
• Fun - Sometimes it's nice to write C

If you have Ruby installed and working, the original is totally fine. This version matches it feature-for-feature.

The Philosophy

Your brain doesn't work in neat folders. You have ideas, you try things, you context-switch like a caffeinated squirrel. This tool embraces that.

Every experiment gets a home. Every home is instantly findable. Your 2am coding sessions are no longer lost to the void.

FAQ

Q: Why not just use cd and ls? A: Because you have 200 directories and can't remember if you called it test-redis, redis-test, or new-redis-thing.

Q: Why not use fzf? A: fzf is great for files. This is specifically for project directories, with time-awareness and auto-creation built in.

Q: Can I use this for real projects? A: You can, but it's designed for experiments. Real projects deserve real names in real locations.

Q: What if I have thousands of experiments? A: First, welcome to the club. Second, it handles it fine - the scoring algorithm ensures relevant stuff stays on top.

Q: Ruby version or C version? A: Both are maintained and feature-complete. Use Ruby if you have it. Use C if you want speed or don't have Ruby. Can't go wrong either way.

Performance

• Startup time: < 5ms (vs ~100ms for Ruby)
• Memory usage: ~2MB (vs ~20MB for Ruby)
• Binary size: ~100KB (vs Ruby + stdlib)
• Fuzzy search: Same algorithm, same results

Both versions are plenty fast for interactive use. The C version just happens to be really fast.

Contributing

Pull requests welcome! This is a pretty straightforward C codebase:

git clone https://github.com/tobi/try-cli.git
cd try-cli
make          # Build
make test     # Run tests
./dist/try    # Try it out

See CLAUDE.md for architecture details and development guidelines.

License

MIT - Do whatever you want with it.

Credits

• Original concept and Ruby implementation: Tobias Lütke
• C port maintains feature parity with the Ruby version
• Uses z-libs for safe string/vector handling

---

Built for developers with ADHD by developers with ADHD.

Your experiments deserve a home. 🏠