Agent Mode
I’ll try that. I just wanted to emphasize that what I’m asking about is a different kind of agent than what I suspect the agent channel is for. My understanding is that the agent channel is for the Convex AI agent component made by Ian, while I’m asking about the agent mode flag you can enable to run Convex without being signed up. https://docs.convex.dev/cli/agent-mode
That way, tools like Cursor’s background agent or OpenAI Codex can set up Convex on their local VM while you do async/LLM coding. So they’re really two different things.
Do you still think Ian and the agent channel is the right place to ask?
7 Replies
hey
@ballingt is the one who added that capability for local development
hey
tnx.
sure. he's familiar with the use cases & etc
thought it would be best to go direct to the source.
yeah, i guess you're going after it.
as soon as codex came out in may, i was banging my head against the wall, as i wasnt able to get convex to run in their ubuntu vm.
cool. we work with the codex team at openai a bit, so they've been trying it too
we know more about this stuff now than we did a few months ago. hopefully we can help
i've been trying to figure out ways to solve this, so love your efforts on it. have even considered signing up for another user, just to have two local cloud env.
i know this is an issue that will be solved with time, i'm just inpatient.
i was super spectical about background agent in the beginning, but is now starting to see more uses cases for it. eg, to be able to spin up 2 instances to try to solve something with 2 different models at the same time.
@ZeroFive what are you running into with agent mode? We fixed it to work with Codex a couple weekends ago, it should work as long as you're logged out