Cursor-agent some questions/suggestions

Feature Requests
1

Feature request for product/service

Cursor CLI

Describe the request

I’ll keep this short and sweet.

  1. I’ve noticed in the usage logs that my calls to cursor-agent are all ‘non-max’ calls. I’d personally like to use Max in the CLI too. All the power please.
  2. model selection - is there a specific reason why the models are limited? One of the pros of cursor is model-diversity. It would be nice to have the choice of using sonic or grok-4 or even gemini-2.5-pro if I’m feeling desperate and want an AI model that will tell me that it’s having an existential crisis as it’s just realised it doesn’t have hands and can’t really type…
  3. customisation - simple feature: turn the .cursor/rules into slash commands. It’s very convenient in a CLI to just be able to type /review and that immediately imports the slash command. It’s just file import in disguise.
  4. more customisation - ‘agents’ as claude code likes to call them can just be .cursor/rules that start with agent-*. If this would translate to the IDE it would be pretty fantastic as you’re no longer tied to having to switch around manually the agents, but you can quite literally have cursor spawning the CLI agent and prepend /agent-the-tester-that-never-hardcodes-any-values-or-the-tests-are-useless and it just triggers the sub-agent.

Some ideas that I will hack away now I’ve put them to paper but would rather they were native. Get everyone out of claude-code, and move on to the better option.

3 Likes
2

yes, agree, we defenetly need more models in cursor-cli

1 Like
3

Given Cursor CLI’s potentional for automation (in headless mode) using at least one no-cost OS model like deepseek-v3.1 would be great.

4

There’s no such thing as a “no-cost” model. Just because the weights are open source, someone still has to pay for model hosting, runtime inference costs, and so on.