AI Developer CLI
Your AI CLI,
built with
restraint.
Sakura is in preview. Get notified when public access opens and be first in line to try plan-first AI workflows built for serious developer work.
Capabilities
Structure and safety.
The defaults that matter.
Plan before you mutate
Every operation that changes state generates a plan first. Review what will run, the risk level, and the exact commands — then type APPLY once you're ready.
Interactive chat with approval
Sakura suggests commands during conversation and asks before running them. Allow once, deny, or trust for the session — you stay in control.
AWS-native passthrough
Read-only AWS commands run immediately. Mutating operations require --apply and confirmation. Your existing AWS CLI handles auth — Sakura just adds the safety layer.
No API keys needed
Sign in with GitHub or email. Sakura handles the AI providers on the backend — you never configure or manage API keys. Just install, log in, and go.
Workflow
From intent to applied,
every step visible.
$ curl -fsSL https://cli.sakura-ai.dev/install | bash $ sakura login --github ✓ Logged in as dev@company.com
Install and log in
One command to install. Sign in with GitHub or email — no API keys to configure. Sakura handles the AI providers on the backend.
$ sakura plan "scale API to 4 replicas" Route → aws · mutating Summary → Update ECS desired count to 4 Saved → .sakura/plans/latest.json
Plan before you act
Describe what you want in plain language. Sakura classifies your intent, generates a structured plan, and shows the risk level — nothing runs until you say so.
$ sakura apply Type APPLY to confirm: APPLY ✓ Plan applied successfully
Apply with confidence
Review the plan, then confirm. Sakura executes exactly what was reviewed. Full provenance is recorded — provider, model, timestamp, and commands run.
Why Sakura
Built differently,
by design.
Most AI tools automate noise. Sakura adds structure to signal. It's designed for developers who want AI assistance without surrendering control or visibility.
Core guarantee
No mutation reaches your infrastructure
without an explicit plan and confirmation.
Explicit over implicit
You review the plan. You type APPLY. No tool should make consequential decisions about your infrastructure without your input.
Calm over clever
Sakura doesn't try to impress you. It plans concisely, asks before it runs, and executes exactly what was reviewed — without elaborate ceremony.
Structure over trust
We don't ask you to trust the AI. We ask you to read the plan, check the risk level, and confirm. The gate is the feature.
Provenance by default
Every response is tagged with model, provider, and timestamp. You always know what ran, who ran it, and what plan it was attached to.
Get started
One command to install.
Install Sakura, log in, and start building. No API keys to manage — everything is handled through your account.
$ curl -fsSL https://cli.sakura-ai.dev/install | bashRequires Node.js 20+
Quick start
Log in to your account
Sign in with GitHub or email. No API keys needed.
$ sakura login --githubInitialize your project
Creates .sakura/config.toml and a local plans directory.
$ sakura initStart building
Chat interactively or generate your first plan.
$ sakura chatRequires Node.js 20+. Sign in with GitHub or email after installing.