AI work needs a clean local stack
Working with AI means your PC needs a solid DevOps environment — not just a model chat window. Agents, Claude Code / Codex-style harnesses, GitHub & Firebase deploys, and data CLIs like Supabase, Databricks, or Snowflake all depend on tools being installed, on PATH, and verified. Sprout is the starter that installs and fixes those tools so your machine is ready for AI work.
-
Agents & harnesses
Bootstrap and repair local tooling for coding agents and Claude Code / Codex-style harnesses.
-
Deploy CLIs
Get GitHub, Firebase, and other ship-to-prod CLIs installed cleanly for simple DevOps flows.
-
Data tools
Install and fix database and data-platform CLIs — Supabase, Databricks, Snowflake, and more.
Three steps to your first install
Install Sprout, add your Pollinations key, then install the tools your AI workspace needs. Sprout detects your machine before it changes anything.
- 1 Install
- 2 API key
- 3 First run
Install
Install from npm, use the GitHub release tarball fallback, or build from source.
Release tarball
From source
npm install -g sprout-install
npm install -g https://github.com/murderszn/sprout/releases/download/v0.3.0/sprout-install-0.3.0.tgz
git clone https://github.com/murderszn/sprout.git && cd sprout && npm install && npm run build && npm link
Do not install from murderszn.github.io/sprout; that URL is the website, not the package.
API key
Run sprout login to authorize with your Pollinations account (BYOP — uses your Pollen balance). Or paste an existing sk_ key with sprout config --set-key.
sprout login
Authorized keys save to ~/.sprout/config.json (chmod 600). Alternatively: sprout config --set-key or export SPROUT_API_KEY.
First run
Install a tool, diagnose a failed attempt, or dry-run the plan with zero side effects.
sprout install gh
sprout --dry-run install node
What you run
Sprout's whole world is the local tools AI work depends on — package managers, PATH, and shell config — nothing else.
// INTERACTIVE FEATURE DEMOS
01. App Installation
Install the Sprout runner globally from the npm registry to get the CLI tool ready on your machine.
02. Tool Installation
Main installation flow: detects your OS, builds a step-by-step plan, asks permission, and verifies.
03. Dry Run Simulation
Preview the installation sequence with zero side effects. No files or system paths are modified.
04. PATH & Environment Repair
Paste failed build logs or shell errors to diagnose path conflicts and automatically repair shell configurations.
Detect → plan → confirm → verify
A narrow CLI agent — not a general coding assistant. It orchestrates brew, apt, npm, pip, and friends instead of reimplementing them.
Read this machine
OS, shell, rc file, architecture, package managers, PATH — one typed snapshot everything downstream uses.
Plain English first
The model states what it'll check, install, and verify — before any command runs.
You approve each step
Every command shows as an argv array with a reason. Sudo, system package managers, and rc edits ask [y/N] — default No.
Prove it worked
Sprout runs the tool's verify command itself and shows real output. Done means the verify passed.
Hard guardrails
No curl | bash, ever. Destructive patterns are blocked before confirmation — --yes can't override them.
One job only
Won't write your code, review PRs, or answer unrelated questions. Install and repair tooling — that's it.
Bring your own Pollinations key
Sprout never ships or proxies an API key. Inference goes to
Pollinations via its OpenAI-compatible API —
your sk_ key, stored in ~/.sprout/config.json (chmod 600)
or exported as SPROUT_API_KEY.
Curated install recipes
Per-OS install and verify commands for common tooling and modern agentic coding CLIs — the agent adapts when reality doesn't match the script.
Common questions
Straight answers before you run anything on your machine.
Do I need an API key?
Yes — inference uses your Pollinations account. Run sprout login (recommended; BYOP device flow at enter.pollinations.ai) or paste an sk_ key via sprout config --set-key. Use sprout login --no-save for session-only authorization, or sprout config --clear-key to remove stored credentials. Sprout never ships or proxies a key; yours stays local.
Is it safe to run on my machine?
Every step shows the exact command and asks for confirmation before sudo, system package managers, or shell rc edits. Destructive patterns are hard-blocked — --yes cannot override them. No curl | bash, ever.
How is this different from ChatGPT or Cursor?
Those are general coding assistants. Sprout only handles installs, PATH, and shell config. It detects your OS, shell, and package managers, runs real verify commands, and refuses unrelated questions.
What if I don't use Homebrew?
Sprout detects what's actually on your machine — apt, dnf, npm, pip, and others. It picks a path for your setup and tells you when it's reasoning outside the curated knowledge base.
My install already failed — can Sprout help?
Yes. Pipe the log: brew install foo 2>&1 | sprout diagnose --tool foo or run sprout diagnose and paste the output. It proposes a fix plan the same way as a fresh install.
Does it work on Apple Silicon, Linux, and Windows?
v1 targets macOS, Linux (including arm64), and Windows. Run sprout env to see the snapshot Sprout uses for your machine.
What does it cost?
Sprout is MIT-licensed and free. You pay Pollinations for inference from your Pollen balance (sprout login) or your own API key. Installing Sprout has no subscription.