Codex defaults to 3 subagents — one config.toml line raises concurrency to 64

daniel_mac8 · x · 2026-09-07

Codex by default runs at most 3 subagents concurrently, but OpenAI tested its Astra with up to 64.

Add this to your config.toml:

[agents]

maxconcurrentthreadspersession = 64

The author demos Astra orchestrating 64 Luna subagents in parallel; a follow-up notes the Codex UI becomes laggy, likely from CPU overload.

Related event: Codex subagent concurrency can be raised to 64, at a UI cost(2 posts)→

Original post →

More from coding & agent

coding & agent channel →