Secret broker for coding agentsFor macOS

Your agents use the keys. They never see them.

Claude Code and Codex work with ms://stripe/test. The real key stays on your Mac.

$brew install csakash/tap/morsec
One key, two requestsLive
How it works

Names travel. Values arrive only where they're allowed.

01

Keys become references

Your .env keeps names. The keys move to your keychain.

02

Swapped in on the way out

The key is added on the way out, only for hosts you allow.

03

Anywhere else, it stays a name

Sent anywhere else, it stays a name. Even a tricked agent leaks nothing.

04

Production waits for you

Production keys wait for your Touch ID.

Security model

Built for an agent that runs as you.

read the key

Locked to morsec

Only the signed morsec daemon can open your keychain items.

approve itself

You approve, not the agent

Approvals happen in your terminal, with Touch ID.

cover its tracks

Every use is recorded

Allowed or blocked, each use leaves a tamper-evident receipt.

By the numbers
0

secrets in the agent's context

3

ways to install

1

command to cut every agent off

Rolling out to a team?

Let's set up your agents.

Book a discovery call
Get started

Set up in five minutes.

~/code/shop
1$ brew install csakash/tap/morsec && ms init
2$ ms import .env
3$ ms rules stripe/test --add-host api.stripe.com
4$ ms connect claude-code
5$ ms doctor
Questions
01Does morsec send my secrets anywhere?

No. Keys stay in your Mac's keychain and only go where you allow.

02Which agents does it work with?

Claude Code and Codex today, plus any MCP client. Cursor and Gemini are next.

03What if the agent edits its settings to skip morsec?

Nothing leaks: it never had a key. It just loses access.

04What happens when msd isn't running?

Nothing goes out. morsec fails closed.