shll

$ ls tools/

ideahopfab-kitwtrun-kittushllall tools →

# one line, the whole toolkit

curl -fsSL https://shll.ai/install | sh

then wire your shell:

shll shell-setup      # wire your shell integration
run-kit agent-setup   # optional, once per machine
exec $SHELL           # reload your shell

Installs shll and the whole toolkit via Homebrew — see the full install guide.

$ shll install==> [1/6] wt==> [2/6] idea==> [3/6] tu==> [4/6] run-kit==> [5/6] hop==> [6/6] fab-kit==> done in 12.3s$ shll versionshll     v0.0.21   [git]idea     v0.0.15   [git]hop      v0.1.18   [git]fab-kit  v2.15.4   [git]wt       v0.0.23   [git]run-kit  v3.7.0    [git]tu       v0.7.0    [git]$ ▊

---

$ cat ABOUT.md

The shll toolkit is seven small CLIs for AI-assisted coding — independent, brew-installable tools that each do one job well and compose into a single workflow. Together they run the whole loop: from capturing an idea to a merged pull request, with the AI doing the typing and a plan in place before it starts.

Built for developers running multiple coding agents in parallel: every backlog, plan, and cost report is a plain file you can cat, grep, and git diff — the contract lives in the file, not in hidden state. No SDK, no vendor lock-in, nothing hidden from git.

It’s not another coding agent — it’s the workflow layer above the one you already use. shll doesn’t compete with Claude Code, Codex, or Cursor; it organizes and scales how you use them.

$ ls -l tools/

• ideacapture ideas and feed a backlog without breaking flow
• hopa personal directory of your git repos — jump anywhere, batch-update from anywhere
• fab-kitthe planning harness — a constitution and a plan before any agent writes code
• wtdisposable git worktrees so each change works in isolation
• run-kitone dashboard to watch every agent session
• tutrack what your AI coding sessions cost
• shllthe bootstrap — install and update the whole toolkit

# the loop — one idea fans out into parallel agents, converges into one dashboard

One idea fans out into parallel agents — each in its own isolated working copy — and converges back into a single dashboard. Inside the loop, idea captures the backlog, fab-kit turns an item into a plan, wt gives each change a disposable worktree, and run-kit watches every session. Around the loop, hop jumps between repos, tu tracks what it all costs, and shll keeps the toolkit installed.

$ whoami

Sahil Ahuja — I build small tools that make AI coding agents plan before they type. The shll toolkit is the workflow I use every day, packaged. Find me on GitHub, LinkedIn, or at noon.design.

# Questions or feedback? → Join the Discord