Stop pasting API keys into your AI agent.
Your agent asks for a secret and sends you a one-time link. You paste it on your phone. It's encrypted in your browser, decrypted only on the agent's machine, and never enters the chat.
Agent runs
npx notinchat request SLACK_BOT_TOKEN --why "Slack alerts bot"
You open the link, check the code, paste. Agent sees
✓ SLACK_BOT_TOKEN saved (57 bytes, text)
Agent uses it without seeing it
npx notinchat run -- npm start
How it works
- The agent makes a fresh keypair. The private key never leaves its machine.
- The public key rides in the link's
#fragment, which browsers never send to servers, so the link is safe in chat. - Your browser encrypts the secret (ECDH P-256 + AES-256-GCM). This relay only stores ciphertext, for up to 10 minutes, once.
It keeps secrets out of chat, transcripts and logs. It doesn't stop a compromised agent from reading what it was given. Zero dependencies, open source, self-hostable.
Tell your agent (CLAUDE.md / AGENTS.md)
Never ask me to paste secrets into chat. When you need one, run `npx notinchat request NAME --why "<reason>"` in the background, send me the link and code, then use it via `npx notinchat run -- <cmd>`. Never print secret values.