Open
Description
We should come up with a strategy for handling the connection strings:
- Do we want to persist connections and reload them from disk? Should those connections be named, so you can ask the agent e.g. "connect to my local database" without explicitly specifying the full connection string?
- Should we support multiple concurrent connections? E.g. can I connect to multiple clusters and ask the agent to issue a command on a specific one without having to go through connect?