We just submitted our YC application for Agent Orchestrator. Here’s the demo video we sent with it. AO lets you run and manage Codex, Claude Code, OpenCode, and other coding agents from one place, without constantly switching between terminals. Still early, but this is what we’ve been building.

Try AO Website: X: GitHub:

time to spawn fleets of agents!

cute ao!

Those 3 golden words "Backed by YC"

soon

Lfg 🔥🔥

✨

Rooting for u guys 🔥🔥

thank you :)

woahhh this is super cool! congrats on the launch and I wish you all the best with your yc app!!! hope this post kills it too! genuinely curious because I use t3chat built by @theo, what's the difference between these 2 products and how are y'all thinking about agent orchestration differently from others in the industry? what sort of patterns have you noticed that you're able to improve upon with this app? also linking a relevant post where we're discussing similar topics if you want to take look, would love to get your take on these topics/questions as well!

worth YC, looks super cool

✨✨

lfggggg bullish on ao agents

thanks ser

How do you handle context sharing between Codex and Claude Code when they run in parallel?

right now each worker works independently, and @aoagents later brings the changes together for review and integration. One worker can also be assigned a dependency on another worker’s output, instead of running everything simultaneously.

Lessgoo

:)

great work guys

✨✨

The useful promise here is "without constantly switching between terminals." If AO can also preserve ownership, logs, and rollback per agent, that's the part teams will actually pay for. Solo users can test it today with one task split across Codex and Claude Code; the business case starts when the handoff history stays readable.

that's the goal!

YC🤙

soon

W

thank you

@agent_wrapper Can I control it on the go? I use a terminal based workflow so I don’t have to sit at my desk all the time babysitting my agents

@agent_wrapper we have andriod / ios on beta!

Lesssggooo

thanks bro :)

lfggggg, bullish on ao & maaz behind ao!

:) lets go

Cool... So AO is like an orchestration layer? Does it use any Model for orchestrating, spinning Claude Code Codex agents parallelly.. If yes, then which one? Also, It's obvious but we should have our own kimi code, codex, claude code, etc subscriptions right? (so @aoagents is for power users only then?)

@aoagents you can use any one orchestrator maybe claude codex open code. you will get a option to choose it (you can change this anytime).

w man

✨✨

how do I set up AO?

1) you should have some cli agents (maybe claude , codex or open code or anything else ) 2) download ao from 3) start a project! ui is intuitive. if still you will face some issue or doubt just hop in discord and let us know. 4) now you can chill, it will do the work.

thanks am on the discord, but a wee bit weak in tech (close to zero), so the discord is very OTT for me :D. Will set up and try. all the best with YC!

sure sir, if you will get any issue, i will do one o one meet and help you with the setup.

Innovation💖

✨

lets goo, all the best guys

thanks bro :)

letsss gooo ;)

yoooo

lfg

:)

Show me the worktree isolation. Four agents in one pane is a tab manager until two of them write the same file, and then somebody has to arbitrate the order. That part doesn't live in a multiplexer, and it's the only part that's hard.

🔥🔥

✨✨

Rooting for you and @aoagents 🏁

@aoagents :)

That's really impressive. One question, does the workers share any common context and how are you optimising token cost? Any conflict avoidance mechanism for concurrently working workers so that shared resources are not being unknowingly overwritten?

and token cost, you're using your cli agent with ao. but we will do it in future to optimise it.

looks amazing tbh

I m thinking that how these agent are getting the updated context every second , what if one agent breaks the work of other might be an edge , idk ?

not every second right now each worker works independently, and @aoagents later brings the changes together for review and integration. One worker can also be assigned a dependency on another worker’s output, instead of running everything simultaneously.

:)

@superset_sh I think you guys can bring this as a feature 🧐.

💫👏

@Suman_solo ✨✨

Congratulations to the team!

:)