Move the plaintext secrets in your dev setup into a 1claw local vault, and let your AI coding agent use them without ever seeing the values.
1claw setup connects your agent to the vault. secure-my-setup fills the vault, safely, from the mess you already have, then hands off to that same secure wiring.
- Scans your project (and, with
--global, your machine) for plaintext secrets. - Moves them into your 1claw local vault, verifying each one landed before it touches a file.
- Strips them from your files, gitignores those files, and flags any that were committed to git for rotation.
- Wires your AI harness (Cursor, Claude Code, Codex, Hermes) to the local daemon, so the agent fetches secrets through 1claw's guarded proxy instead of reading them in plaintext.
- Node 20 or newer.
- The 1claw CLI:
npm i -g @1claw/cli(orbrew install 1clawAI/tap/oneclaw).
# dry run: see what would happen, nothing changes
npx secure-my-setup
# move secrets into the vault, strip them from files, wire your agents
npx secure-my-setup apply
# run your app with the migrated secrets injected into its env, nothing on disk
npx secure-my-setup run -- npm start
# put the secrets back into your files
npx secure-my-setup undoAdd --global to also scan ~/.zshrc, ~/.aws/credentials, and ~/.config. Add --json for machine-readable output.
secure-my-setup never sees your vault passphrase, never writes a credential into a harness config, and never sends a secret over the network. The secure wiring uses 1claw's local daemon: the agent calls a proxy that injects the secret into outbound requests server-side, so the raw value never enters the model's context.
See SECURITY.md for the threat model, and scripts/live-proof.sh for a runnable proof against the real daemon.
npm install
npm run build
npm testMIT