Credentials.
Agents use real accounts. Keys stay out of the chat.
Tokens and API keys are stored encrypted, entered through a form, and applied when a tool or a command needs them.
Stored once, encrypted.
Encrypted at rest
Tokens and API keys that Kylon stores are encrypted at rest.
Entered through a form
Agents ask for a key through a secure card instead of the chat. An existing key can be rotated, added or renamed in place, on the same connection.
Not shown back
The app does not display a stored key again. A person who manages the connection can copy it when they need it.
Applied when used.
Connected tools run on Kylon's side
When an agent uses a connected tool, Kylon applies the stored credential and returns the result to the agent.
Checked before every command
When an agent's own command needs a key, access is checked again before the command starts, and the key is provided to that command rather than written into the conversation.
On your own computer
For agents running on your computer, the Kylon CLI passes one connection's keys into one command and removes exact copies of them from its output.
On the surface
A Connect button, a form to add or rotate a key, and no keys in the conversation.
Keys end up pasted into prompts, chat logs and memory, and nobody can say which agent holds what.
One place to add and rotate keys, and agents that use them without asking for a password in the chat.
Good
questions.
Can an agent see my key?
A connected tool never hands the key to the agent; Kylon applies it and returns the result. A command that needs the key itself receives it while it runs, so grant API keys to agents you trust with that service.
What happens when I remove access?
New tool calls and new commands are checked again, so they stop using the key. A command that already received it keeps it until it ends, so rotate the key if it may have been copied.