Hit your usage limit in Codex/Claude Code mid-task? Orca now lets you resume the exact same session in a different agent harness — one click. Codex ⇄ Claude Code, full context intact🔥

Gat started:

i mean...

Context intact is doing a lot of work here. Transcript transfer is easy. Preserving tool state and unfinished plans across harnesses is the real test.

Woahhh this is dope

@JinjingLiang Incredible tech 🔥

@JinjingLiang fire 🔥

@JinjingLiang This is 🔥

@JinjingLiang 🫶

Wow awesome 🎉 I usable just use entire and have my next agent pick up where the last 1 left off. But this seems even easier, definitely going to have to try it

🔥 but could u optimize the git section :(

its like when your uber cancels but ur already in the car 💀

It should run ce-handoff 😀

Great work, this project looks really cool I'm going to check it out!

So good! Are you thinking about expanding the agent panel to a full Kanban Board?

👀 some experimental agent-centric view is in-progress

Best ADE ever

@JinjingLiang You're the best 🔥

This feature is great, but I'm having a lot of trouble when I have many sessions. The names in the session list don't match the custom names I've set in OC, which is causing a lot of confusion.

That makes sense. Agree the agent session name can be improved to help with locating them. Filed to track this work!

Last time I ran out of context window was in 2025

always shipping exactly what we need

Wow 💪

isnt this what @opencode has done for the last year lol

11 PRs in 17 hours from one small UX tweak is wild.

i can do better implementation of it.

Same as Alex, but you can also prevent the limit by automatic transparent failover.

codex sending a notification for every mcp (and maby all other tools) tool calls

Crazy 🔥🔥

"Full context intact" is the expensive part. I audited my own transcripts: one session sitting at ~400K context cost over $1 per turn to emit 470 tokens. Your Focused handoff default is the better one. Transferring is cheap. Every turn afterwards re-reads the lot.

Useful when you hit a limit mid-task. From doing this by hand across harnesses: the transcript is the easy half. What breaks continuity is the reasoning that never got written down — which approach was already tried and rejected, and why. The new harness inherits the tokens, feels confident, and walks straight back into the dead end. Do you carry anything structured across (decisions, rejected paths), or is it the raw context window?

curious how you're serializing the context across harnesses, thats the gnarly part the tool call history alone can get messy between different agent formats

this solves a real thing. hitting a limit mid-task and losing all the context just to switch tools is the worst part of running multiple harnesses. portable session state is the feature nobody thinks they need until the exact moment they do

Great feature! How would this work if trying to start a new session from 1M context Claude session to 272K context Codex session? Also this has potential implications on burning more tokens transferring huge context over to a new session.

This is live?

yes!