1 (not long running)

I run multiple fable/opus and sol sessions for both chat and coding- I read a lot of academic papers and it grew to a point where I needed to use a reference management system like zotero to keep track of it. So I recently one-shotted a ~1000 pdf literature review folder into a .bibtex file to import into zotero. . Includes multi-step DOI extraction and multi-review steps to reduce false positives. Also can check patents via EPO API. And can interact programmatically with zotero via its API to add tags, notes etc. Very specific to my requirements and codex sol max did it in a single 8-hr goal session with very few bugs

Depends on how you define agents. I use CodeRabbit all the time but it has a swarm of dozens of sub-agents. So am i using 1 agent or 10+?

I'm obviously biased, but I'm building aq.dev, a shared workspace for running coding agents with a team, so I watch a lot of people's sessions more than my own haha.

If I'm between meetings, then I'm really watching 1-2 max. There could be a few others running in the background, but those I'm not paying attention to that much.

If I'm coding and in the zone, then usually 4-5 workspaces with 1-3 agents per workspace.

I have a tmux split in 4 panes, 4 agents running in 4 workspaces with shared memory. Most of the times I use 2 concurrently, but here and there I have extra analysis or reviews to do. I like this setup with tmux, works over ssh, easy to bring any one full focus, etc.

6-8 running nearly 24/7, any more than that and there's too much for me to review

whatever codex spawns and uses

I rarely call multiple AI agents at the same time, because I often get the context wrong (such as talking about the needs of project B in project A), and I feel that the parallel ability of the human brain is very limited.

At least 4

I just use 1... been using Kimi and it’s been great.

I use Codex, which defaults to 1 agent and automatically dispatches more when handling complex tasks. That's more than enough for me