Envlet for Codex

Codex reads its MCP servers from config.toml. One block adds Envlet, one token gives it access.

envlet run -- codexproduction withheld

Set up

Three steps. The host keeps one secret and everything else arrives at boot.

Add the block to ~/.codex/config.toml.

~/.codex/config.toml

[mcp_servers.envlet]
command = "npx"
args = ["-y", "@envlet/cli", "mcp"]

02
Give it a token
Add Codex under Access in the dashboard, or let the migrate prompt do it. Its token opens one environment, and production starts fully withheld.
agent/codexagent
Its own token, separate from your login.
token
envlet_•••••••• shown once
opens
one project, one environment
expires
always
production
withheld
03
Run it through Envlet
Start Codex with envlet run. It sees the values its token allows and nothing else, and every read shows up under its name.
Terminal
envlet run -- codex

The migrate prompt runs the whole migration in Codex. Values never enter the chat.

Codexenvlet mcp

Move us to Envlet.

01Signs you in if it has to. You approve in the browser.
02Creates the project with development, staging, and production.
03Finds every .env file and imports each one. No value enters the chat.
04Links the repository so every command knows its environment.
05Creates a token for your CI provider.
06Creates its own token. Production stays withheld until you name what it may read.
07Puts envlet run in front of your start scripts.
08Lists the names you can now delete from .env files, GitHub, Vercel, and AWS. Names only.

Can Codex read production secrets?

Not by default. A new agent grant withholds every current and future production variable. You allow specific names when you decide to.

Do agents count toward my plan?

No. Agents and CI jobs are unlimited on every plan. Only humans are members.

One command. The right values for whoever runs it.