Hey, really enjoying AgentPlex so far. The multi-session canvas is a great way to work.
One thing I noticed while using it: every Claude session starts with the same flags. It would be handy to be able to set CLI flags per session when creating one.
Why this would be useful
Different sessions tend to have different needs. For example:
- A research session works well with standard Allow/Deny prompts
- A trusted batch task in an isolated folder might benefit from
--dangerously-skip-permissions
- Being able to set
--model, --max-turns, or --allowedTools per session would add flexibility
One possible approach
An optional "CLI flags" field in the New Session dialog. Whatever the user enters gets appended to the command written to the PTY.
For sessions running with elevated permissions, some kind of visual indicator on the node (a badge or border colour) would help keep track of which sessions are running without prompts. Feels especially important when you have a few sessions going at once.
Happy to help
If this is something you would consider, happy to contribute or test. Just thought it was worth raising as it feels like a natural fit for how AgentPlex works.
Hey, really enjoying AgentPlex so far. The multi-session canvas is a great way to work.
One thing I noticed while using it: every Claude session starts with the same flags. It would be handy to be able to set CLI flags per session when creating one.
Why this would be useful
Different sessions tend to have different needs. For example:
--dangerously-skip-permissions--model,--max-turns, or--allowedToolsper session would add flexibilityOne possible approach
An optional "CLI flags" field in the New Session dialog. Whatever the user enters gets appended to the command written to the PTY.
For sessions running with elevated permissions, some kind of visual indicator on the node (a badge or border colour) would help keep track of which sessions are running without prompts. Feels especially important when you have a few sessions going at once.
Happy to help
If this is something you would consider, happy to contribute or test. Just thought it was worth raising as it feels like a natural fit for how AgentPlex works.