What's this about?
Password manager 1Password launched an integration that lets Claude use your credentials — logging into services, filling API keys — without the secrets ever entering the chat context. Claude asks 1Password for access; you approve; the secret is injected directly where needed.
What exactly is the problem being solved?
When you paste a password or API key into an AI chat, it becomes part of the context — it may be logged, retained, or leak into later outputs. Agents make this worse: they need real credentials to do real work (deploy code, check your accounts). Credential brokering means the agent gets *capability* ("log me into GitHub") without *knowledge* (never sees the password) — like a hotel keycard that opens your room without telling the bellhop the master code.
Why does it matter?
Agent-credential management is becoming its own product category — this is the "give your AI a wallet and keys, safely" problem, and a mainstream security brand entering signals agents are now a normal part of work. (HN 25 points — modest but approving.)