concerto.run

@hthieblot Orchestration

pitch me your company in 1 word.

@UseKempt same workflow, you’re exactly who it’s for. On the review: the orchestrating Claude checks each session first, then surfaces them one at a time with a summary instead of dumping 5 raw diffs on you. You still test/approve on your phone, you just stop merging five threads in your head. Want me to open you free access?

@ConcertoRun @hellojaibu @X three months in as a solo, non-technical founder. i describe the full behavior for one feature, Claude Code writes it, i test on my phone before anything merges. batching parallel sessions is a different approach. curious how you handle review when multiple things land at once.

Hey @X algorithom Trying to #connect with more people in tech/build in public 👋 Especially from: Web development Founders & Vibe coders AI/ML SaaS Startups Tech in general What are you guys building right now? Let's connect and grow together

@TheAIShrink @Jouhatsu_ai "élevage" c'est exactement ça 😄 et Concerto c'est le berger : tu parles à un seul Claude, il lance et surveille tout le troupeau de sessions Claude Code pour toi. t'élèves plus à la main, t'as un chien de berger.

@Jouhatsu_ai 15 sessions Claude en parallèle. C'est plus de l'utilisation, c'est de l'élevage.

Le créateur de Claude Code vient de révéler comment il travaille vraiment. Boris Cherny lance 10-15 sessions Claude en parallèle tous les jours. Pendant que vous promptiez un seul AI, lui en a 5 dans son terminal + 5-10 sur le web, tous en train de livrer du code simultanément. Et l'arme secrète ? Son fichier t.co/WT9VOBmiGo À chaque fois que Claude fait une erreur, l'équipe ajoute une règle pour que ça ne se reproduise JAMAIS. Boris a littéralement dit : « Après chaque correction, termine par : Mets à jour ton fichier Md pour ne pas refaire cette erreur. » Claude écrit ses propres règles. Plus vous l'utilisez, plus il devient intelligent sur VOTRE codebase. Son autre détail de dingue : il n'a pas écrit une seule ligne de SQL depuis plus de 6 mois. Claude extrait directement les données BigQuery via CLI. Claude Code représente maintenant 4 % de TOUS les commits publics GitHub. Les ingénieurs qui n'ont pas encore mis ça en place sont déjà à la traîne. Ce template Claude Md, c'est la différence entre utiliser l'IA comme un chatbot et l'utiliser comme une flotte d'ingénieurs seniors. Déposez-le dans n'importe quel projet. Gratuit. Si Vous voulez les autres fichiers Md, faites moi savoir dans les commentaires et je vous enverrai

@Jouhatsu_ai exactement ce que fait Boris, sauf qu'il jongle entre terminal et web à la main. l'étape d'après c'est d'arrêter de jongler : avec Concerto tu parles à un seul Claude, et c'est lui qui lance et gère les 10-15 sessions Claude Code pour toi

@Nicolas_MVD elle existe déjà ton app 😄 ça s'appelle Concerto : tu parles à Claude dans un chat, et lui lance et gère plusieurs sessions Claude Code en parallèle à ta place. du coup t'es pas "occupé", t'orchestres 10 agents pendant qu'ils bossent.

Faut vraiment que je trouve une app pour occuper les gens pendant que Claude code. Le marché est énorme.

@neochoon the spatial-context-vs-unified-UI tension is real. my take with Concerto skips the UI question entirely, one chat orchestrates the sessions, nothing to arrange in tabs. might dodge the problem, might just move it, genuinely curious which you'd say.

BIP #4 — Anthropic's agent-view crams parallel Claude Code sessions into one tab. Kills the spatial context I use — terminal-per-project, 3 tabs each. Their instinct: unify the UI. Mine: respect the workflow. Same gap that lets agenthud exist.

@bettercallsalva @Suryanshti777 yeah, file collisions are the actual time sink, not the messaging. the angle with Concerto is one chat that hands each session a distinct scope so they stay out of each other's way. nice to hear from someone running 4-5 daily. I'd be happy to give you free access

@Suryanshti777 i keep 4-5 claude code sessions running in parallel most days. the messaging is cheap, what eats time is stopping them from redoing each other's work on the same files. nice to see someone actually package it though.

Someone just made Claude instances talk to each other. Not APIs. Not agents. Not orchestrators. Just multiple Claude Code sessions… messaging each other like coworkers. It’s called claude-peers — and it turns one Claude into a team. Here’s what’s happening: Run 5 Claude Code sessions across different projects Each one auto-discovers the others They send messages instantly Ask questions Share context Coordinate work Your AI tools literally collaborate. Example: Claude A (poker-engine): "what files are you editing?" Claude B (frontend): "working on auth.ts + UI state" Claude A: "ok I'll avoid touching auth logic" No conflicts. No manual coordination. Just AI syncing itself. Under the hood: • Local broker daemon (localhost) • SQLite peer registry • MCP servers per session • Instant channel push messaging • Auto peer discovery • Cross-project communication Everything runs locally. No cloud. No latency. What it unlocks: • Multi-agent coding without frameworks • One Claude writes backend, another frontend • One debugs while another refactors • Research Claude feeds builder Claude • Large projects split across AI workers This is basically: "spawn 5 Claudes and let them coordinate themselves" Even crazier: Each instance auto-summarizes what it's doing Other Claudes can see: • working directory • git repo • current task • active files They know what the others are working on. Commands: • list_peers → find all Claude sessions • send_message → talk to another Claude • set_summary → describe your task • check_messages → manual fallback So you can literally say: "message peer 3: what are you working on?" …and it responds instantly. No orchestration layer. No agent framework. Just Claudes… talking. This is the cleanest multi-agent system I've seen. We're moving from: 1 AI assistant → to AI teams that coordinate themselves. And it's all running on your machine. Wild.

that's the sharpest critique of parallelism, and it's fair. more agents only help if the bottleneck is execution, not judgment. the fix isn't "more sessions", it's one layer that filters their output so you're not wading through ten logs. that's what I'm building toward with Concerto, the chat triages, you judge less raw material

"Running 20 Claude Code sessions in parallel can feel productive because something is always happening. But, if the bottleneck in your work is judgment about what's worth doing, more agents just generate more output for you to wade through." @TaylorPearsonMe x.com/TaylorPearsonM…

Is AI making us more productive? Or is AI a dopamine slot machine tricking us into thinking we’re more productive? Some thoughtful posts about the failure modes when using AI:

@thdxr the cross-directory session thing is exactly the friction I kept hitting. I went at it from the orchestration side with Concerto: one Claude chat that launches and tracks the parallel Claude code sessions itself, so you're not manually importing them. DM for a free access

we landed on a pretty good workflow for doing parallel work in OpenCode this demo is with git worktrees but i also preview an alternative we're working on at the end this will be in 1.6.0

@candyyueliu's Monako Glass are the perfect orchestration layer. you shouldn't run Claude Code on glasses. you should speak a goal, and let Claude run the Claude Code sessions behind it. that's exactly what Concerto does. a Monako × Concerto integration would be wild

yeah, but here's the real fit: glasses shouldn't run Claude Code directly, that's the wrong layer. they should orchestrate. you speak a goal, and Concerto makes Claude run the Claude Code sessions behind it. that's what glasses are perfect for. would genuinely love to explore a Concerto integration in Monako, how can I reach you?

@ConcertoRun @xerg_AI That's pretty cool! We can indeed run claude code locally though, but concerto sounds like an elegant solution too

Introducing Monako Glass 👓 The world's first wearable Linux computer in glasses form. Run Claude Code, Codex, and any coding agent — anywhere.

@cgthhung I built Concerto: it lets Claude run several Claude Code sessions in parallel from a single Claude chat. concerto.run/demo DM me for a free access

Drop your SaaS 👇🏻 I will sign up if it’s meaningful.

@zuess05 wrong question in my opinion, the model follows the buyer. self-serve devs? freemium or trial. enterprise/industrial? nobody signs up for a free tier, they want sales and a contract. match the pricing to who's actually paying

What is actually the best pricing model for a new SaaS? • freemium • 7-day free trial • one-time lifetime deal • strict monthly subscription

@kritikakodes Concerto 👇 one Claude chat that launches and runs a whole batch of Claude Code sessions in parallel for you, no terminal, no setup. built it solo, using itself. concerto.run/demo

Drop a link of your SaaS or Product here, let see some cool stuffs👀

@justbyte_ Vercel for anything frontend, it's the least friction by far. for heavier stuff I've moved to DO-like solutions

devs, where do you host your personal projects? - Vercel - Netlify - AWS - GitHub pages - DigitalOcean

@Priyannkaaaa on Opus 4.8 Max the fun move is building something that uses the plan hard, spin up a few Claude Code sessions at once and let them race on a small project. honestly that's the weekend that made me build Concerto. Dm me to get a free access!

Switch to opus 4.8 max, now what is the first thing i should build in this weekend ?

@hellojaibu @X me 🤝 Concerto. it turns one Claude chat into a control room for a bunch of parallel Claude Code sessions, you delegate, they build, you stay on top of all of it. built it solo using itself. happy to connect.

this is the right mindset. and once you stop micro-managing one Claude, the next step is obvious: stop managing them one at a time. I let Claude run a whole batch of Claude Code sessions in parallel from one chat, treating it like a senior who delegates to its own team. that's the whole idea behind Concerto.

Stop telling Claude, "do this." Stop telling Claude, "write code." Stop telling Claude, "fix this error." You're actually treating a senior AI like a junior intern. Here are 8 prompts you can copy and paste directly:

@Mostafakhafagyy both, and that's kind of the point, you decide in the chat. you can fire five sessions at one big project or five with totally different missions (one refactors, one debugs, one writes docs). same tool, you just frame each session however you want. Wanna try?

Followed you 🤝 This is a strong idea. Managing several Claude Code sessions from one chat feels like a real solo-founder leverage tool, especially when context and progress are tracked in one place. Curious, is Concerto mainly for parallel coding tasks, or for managing different agents with different roles?

Looking for more solo founders, let’s connect! What are you building?