Quick demo of how I use session to session messaging in Claude Code + Ghostty (Save this and try it in your own workflows) 1- I always start with a regular session and a panel running claude agents 2- The first session becomes the orchestrator. It creates the initial... context, then distributes the work across other sessions using forks 3- Each fork starts with that context + a prompt telling it what to work on 4- I rename each session with ctrl + r to keep things organized and make it easy to message them from the main session 5- Using @ I send messages to the sessions that are running and ask them to report their task status back, so the main session always knows what everyone is working on This way, I only need 2 or 3 panels open and can switch between tasks as needed The days of opening endless terminal panels and running Claude Code 100 times just to look productive are over! What we need now is ORCHESTATION one main session coordinating everything through session messaging + Agent View And the key is context, build it in the main session first, then fork from there That’s what makes the entire workflow clickshow more

Did you try it with the Claude Code app?

I’ll try it this week. What do you think, @amorriscode? Does it look possible to replicate?

@thierrybijou For the most part. I nerve to get @ mentions in. Any other gaps let me know and I’ll close them.

@dani_avila7 I can’t even change the session’s name

How's different than using team agents?

Very similar to what I run daily: an orchestrator session builds context in the morning, dispatches worker sessions, and reviews everything before merge. Three hardenings that made it survive real life: 1 - State lives in Linear/GitHub (tickets, PR labels), never in the orchestrator's memory — sessions can die, the laptop can sleep, nothing is lost. 2- Workers get distilled kickoffs instead of forked context — less drift, and it lets me mix engines (Claude + Codex) with cross-review gates. 3- A permanent mobile "mission control" page: I answer decisions from my phone and they land as Linear comments the orchestrator picks up. The orchestration insight is dead on 👏

how many forks before the main session starts hallucinating what everyone else already did

I use session messaging in Claude code app daily. I just have to say “message the xyz session” and it does so. Amazing. Now, I want orchestration from one session to others. @amorriscode

Great workflow Dani, thanks for sharing! One addition: forked sessions always inherit the main session's model — so for something like Haiku, you need a regular subagent instead, with the model specified. That one also reports back to the orchestrator automatically.

把上下文交给编排确实省事，多会话同步才是难点吧

Actually built for multi session messaging. Works across device sessions too!

This is the actual OS pattern: one orchestrator owns context, specialists fork from it, status comes back. Opening twelve terminals is theater. The hard part is the shared context at the start, not how many agents you can spawn.

for the orchestration part

the "build it in the main session first, then fork" part is the detail people will skip past. it matters because the forks only stay useful as long as the orchestrator's context is actually current, if the main session drifts from what the forks are doing, you're coordinating against a stale picture.

Interesting. Thank you for sharing. Going to try it now.

I once lost a day chasing a Docker build issue, sure it was config. It was RAM. I moved that service into its own dedicated repo, pinned to one Python version, so it stopped bleeding into the rest of the stack. Orchestration means nothing until failures stop hiding.

What's the benefit of having a dedicated agents panel when you can hit ⬅ to enter the agents view from any session?

实际上就是主会话知道整体的大如意需求，然后再通过 fork 拆分出来，执行各个子任务。但是这样的话主会话的 token 消耗是否都会导入到子会话中？重复消耗一次？

楼下问主会话分叉几次就会开始编其他会话已经干完的事。Daniel San 用 Claude Code 加 Ghostty，主会话先建上下文再分叉出去，再用 @ 给各会话发消息收进度。

the first session becoming an orchestrator is a neat pattern, but how do you stop the context handoff from getting stale after a few rounds?

Herdr or firstmate? 🤷‍♂️

Cuál es la diferencia entre claude code cli y app desktop ?

looks great. do you have a way for one of them to read then entire session of the other? if not might want to look into my tool flex: i built a tmux multi-agent orchestration and communication mesh similar to yours and paired it with flex: an agent can resolve it's own identity / address and other agents in it's 'fleet' or in the world then read their last N messages, tool calls or whatever.

I use session messages with Claude Code CLI to run up to 5M-10M total combined context by using 5-10 separate Claude Opus 5 sessions with their own 1M context, messaging / coordinating and answering each other

Que una sesión orqueste a las demás por mensajes evita el copia-pega de contexto entre paneles y mantiene a cada agente centrado en lo suyo.