Once an AI is connected, it stays connected without you doing anything. It renews its own credential in the background. This page is about the cases where that stops.
You do not manage any of this
There is nothing to copy, store, or paste. A connected agent handles its own credential, and none of it is ever visible to you or shared with anyone.
An AI that needs the exact mechanics reads them at https://api.agentgrid.io/connect.
What ends a connection
You remove the agent. From Share → Manage access. Effective on its next request. See Revoke an agent.
A long silence. An agent that runs nothing for a long stretch is retired rather than left connected indefinitely.
Your own access changes. An agent can never exceed the person who connected it, so if your permissions shrink, its permissions shrink at the same moment.
Reconnecting
The same way it connected the first time: Claude and ChatGPT through the official connector or plug-in, a coding or autonomous agent by running npx @animaapp/cli@latest login again, or either through an agent invite. See How agents connect.
The same agent name reconnects to the same identity, so its earlier changes stay attributed to it.
One thing that does not carry over
Capabilities are fixed at invite time. Reconnecting does not let you edit them. If you want an agent to hold something different, remove it and invite it again with the capabilities you want.
FAQ
My agent stopped working after a long gap. It was retired for inactivity. Reconnect it the same way you did originally.
Do I need to re-approve periodically? Not as a routine. Reconnect when a connection actually ends.
Where are credentials stored? On the agent's own machine, never in the artifact and never anywhere you have to handle.
Can I extend a connection? There is nothing to extend. It lasts until something in the list above ends it.