Gareth MorgansSecurity
Stop pasting API keys into AI chat
It is a familiar shortcut: paste an API key into ChatGPT, Claude, or Cursor so the agent can “just call the API.” It works once. Then the key sits in chat history, model context, screenshots, and exported threads.
That fails audits — and it leaves keys in places you cannot control.
Start free beta · Secrets vault
Why paste is risky
Chat products are not secrets managers. When you paste a key:
It enters the model context for that turn (and often the conversation)
It can land in provider logs, browser history, and shared workspaces
Anyone with access to the thread can copy it
Rotation means hunting every paste site, not flipping one vault entry
If your team treats chat as the place keys live, you do not have secret hygiene. You have hope.
Model context exposure
Prompts, skills, and agent recipes should describe *what* to do. They should not embed credentials. Once a key is in the prompt text or the chat, you cannot reliably prove it stayed out of training logs, support exports, or a colleague’s laptop.
The rule is simple: keys never enter model context.
Controlled substitution instead
You need two jobs separated:
1. Store credentials in a private vault with explicit share
2. Substitute them server-side when an approved agent or HTTP call needs them
Joyst’s secrets vault is private by default. You share a secret with a group or the org when you mean to. Agents and controlled calls can use substituted values without pasting keys into the chat the model sees.
That is not a replacement for AWS Secrets Manager or HashiCorp Vault for infra. It is the AI-asset layer: keys used by skills, agents, and MCP workflows your team actually runs.
What to do this week
1. Stop pasting new keys into chat
2. Rotate anything already pasted into shared threads
3. Move working credentials into the Joyst vault
4. Rewrite prompts and skills so they reference vault-backed secrets, not raw keys
5. Tie access to groups — finance keys stay with finance
Pair this with AI governance when you need permissions and an audit trail on who could see what.
Next steps: How-to: add and use secrets, How-to: controlled substitution, and Agents (secrets in packages).
FAQ
Is this the same as a cloud secrets manager?
Complementary jobs. Infra secrets managers protect production systems. Joyst’s vault is for AI assets — prompts, skills, agents, and MCP workflows — with private-by-default share and controlled substitution.
Can agents still call APIs?
Yes. Substitution happens server-side for controlled calls so the model does not need the raw key in chat.
What if someone already pasted a key?
Rotate it. Treat the chat paste as a leak. Then store the new value in the vault only.