CLI

bun alchemy <command> [file] [options]

Every command operates on an alchemy.run.ts stack file (or a custom entrypoint passed as the positional [file], which must exist) and targets a stage.

Command map

alchemy
├─ deploy [file]        # plan → approve → apply
├─ plan [file]          # preview changes, apply nothing
├─ destroy [file]       # delete every resource in a stage
├─ unsafe nuke [file]   # delete everything a provider can list, tracked or not
│
├─ dev [file]           # hot-reloading development loop
├─ tail [file]          # stream live logs from deployed resources
├─ logs [file]          # fetch past log entries
│
├─ login [file]         # authenticate every provider in your stack
├─ profile show|clear   # inspect or wipe stored credentials
│
├─ state stacks|stages|resources|get|tree|clear   # inspect and manage the state store
│
├─ aws bootstrap        # set up the per-account AWS assets bucket
└─ cloudflare bootstrap|create-token|state logs   # state-store worker, API tokens, its logs

Each command has its own page: deploy, plan, destroy, unsafe nuke, dev, tail, logs, login, profile, state, aws, cloudflare — plus two workflow guides, Adopting Resources and Inspecting State.

Common options

Option	Description
--stage <name>	Stage to target. Falls back to the stage env config, then dev_$USER. Must match [a-z0-9]+([-_a-z0-9]+)* (case-insensitive).
--profile <name>	Auth profile from ~/.alchemy/profiles.json. Defaults to default or $ALCHEMY_PROFILE.
--env-file <path>	File to load environment variables from. Defaults to .env.
--yes	Yes to all prompts.
[file]	Stack entrypoint. Defaults to alchemy.run.ts; must exist.

--profile is genuinely common — deploy, plan, destroy, dev, tail, logs, login, and state all take it.

The interactive TUI

In an interactive terminal, deploy, destroy, and plan render a live Ink TUI: the rendered plan, an arrow-key approval prompt, and per-resource apply progress that unmounts into the final output. Everywhere else the same commands emit plain line-oriented logs.

Environment variables override the detection, in this order:

Variable	Effect
ALCHEMY_PLAIN=1 or ALCHEMY_NO_TUI=1	Force plain output
ALCHEMY_TUI=1	Force the TUI
No TTY, CI=1, or a known agent env var (CLAUDECODE, CLAUDE_CODE_ENTRYPOINT, CURSOR_AGENT, AIDER_MODEL, CODEX_CLI)	Force plain output

The TUI module is lazily imported only when selected, so plain runs never pay the Ink import cost.

Plain mode never prompts — it prints the plan, makes no changes, and reports: Non-interactive terminal detected. Pass --yes to approve, or set ALCHEMY_TUI=1 for the interactive UI. In CI, always pass --yes.

Running these commands in a pipeline? See CI.

Where next

• deploy — plan, approve, and apply changes
• dev — hot-reloading development loop
• login — authenticate your stack’s providers
• state — inspect and manage the state store
• Stages — how environments are isolated
• CI — run the CLI in pipelines