Framer AI AgentsDesign and publish professional sites with AI
Your AI coding agents, visible on one persistent infinite canvas. Spawn role-based workers (builder, QA, reviewer, devops) and an orchestrator runs the build, review, fix, validate loop while you steer. Terminals, browsers, notes and git live beside your agents, so checking on your AI is a glance, not an interrogation. Your whole workspace persists across restarts: every agent, session and pane restored where you left it. Works with Claude Code and Codex, locally. macOS, Linux and Windows.
Free Options
Launch Team / Built With
Hey Product Hunt! 👋 Michael here, maker of AgentGrid.
The true origin: my co-founder Souren built AgentGrid to replicate himself. He was running so many coding agents there was no sane way to manage them all. It took him a MONTH to convince me to use it.
What converted me was restart anxiety.
Every system update meant losing all my Claude Code sessions. AgentGrid saves your entire workspace: every agent, every session, mapped visually on a 2D canvas. Restart, reopen, everything's exactly where you left it. I haven't lost a session since, and honestly we don't talk about that feature enough.
Once I was in, the rest clicked: an infinite canvas where your agents are panes you can watch. You spawn role-based workers (builder, QA, reviewer, devops…), and an orchestrator agent runs the loop: build → review → fix → validate, while you steer and take the final review.
Terminals, browsers, code, git and notes live on the same canvas, so "what is my AI doing?" is a glance, not an interrogation.
The part I love most: AgentGrid launched itself. The gallery images above, the outreach research, and the first draft of this comment were produced by an orchestrator and its workers running on the canvas. While I slept.
Works with Claude Code and Codex today, on macOS, Linux and Windows.
It's free during early access: grab it at https://agentgrid.sh.
And if you show up today and tell us what to build next, you're a founding user in our book: your feedback goes straight onto the roadmap, and there's a founding-user role waiting in our Discord.
I'll be here all day. Tell me what's confusing, what's missing, and what you'd want your agent team to do first. Honest teardowns especially welcome. 🙏
PageAI
@michael_yagudaev congrats on the launch, this looks awesome guys!
@dan_mindru thanks so much Dan :). I am legit excited to see how you use it and what you build with it
@michael_yagudaev @dan_mindru Awesome to see you here Dan!!
AstraPixels
@michael_yagudaev congrats on the launch. Great product!
@nicklaunches thank you so much Nick ❤️. Can't wait to see what you create :)
@michael_yagudaev @nicklaunches awesome to see you here Nick!
@michael_yagudaev 🚀🚀
Composio
Great launch. A lovely UI. I'm so excited to try this out
What's the maximum number of agents you've seen something running in parallel? Do you have any resources for how we can start maximising our agent usage while not lowering the quality due to the human being the limiter?
@shawn_esquivel great questions :). About 300+ for me, we are improving performance now as TUI is terrible rendering pref. Once we do that next week it should be 1000+. MacOS has a 511 file description limit you can increase too with a quick bash command.
Typically, I let the agent decide how many parallel tasks to do, claude & codex are really good at that.
The challenge is making the review process breezy for you, I typically have agents review each other and leave screenshots. Triage the PRs with low LOC count for quick win merges for human review and get those shipped to real customers.
I'll be creating more resources as we go and will talk to you guys at Composio on itegration ideas :)
@shawn_esquivel I am so excited to see you try each one! And thank you for the support!
My workflow has been consistent in having multiple orchestrators working on different problems in different work trees. Each of these orchestrators has multiple adversarial reviewers, where each reviewer is a different model and harness. Sometimes I do Codex grok and Claude workers for a feature that a cloud builder shipped, all orchestrated under the same cloud orchestrator.
The primary purpose of AgentGrid is to remove the human from the loop. Obviously we're all the bottlenecks so this is definitely a solution or at least a step towards that.
the infinite canvas framing for multiple agent terminals is a good fit for how messy multi-agent work actually looks in practice. when two agents on the same canvas need to reference each other's output (one's notes inform the other's next terminal command), is that a manual copy-paste between panes or is there some shared context layer connecting them
@omri_ben_shoham1 AgentGrid exposes tools to agents so they can use such tools to query and feed their context. Think of how agents could use bash to hunt down information. On AgentGrid, they can find the info, connect to it, and truly collaborate
@omri_ben_shoham1 exactly :). Creative process is messy and you need space to think. I literally don't know how to build without AG anymore. Broke it a few times early on and it suck having to go back to single terminal session 🤮
@Michael Yagudaev that "suck having to go back to single terminal session" bit resonates, once you've had parallel context it's hard to go back to one linear thread of thought. @Souren Khetcho the bash-hunt-for-info framing helps too, sounds less like separate agents and more like one team with shared eyes on the same canvas. does the shared context layer show which agent found what, or does it get flattened once it's fed back in?
@Souren @omri_ben_shoham1 Thanks Omri :). It depends what you mean "what agent found what". You can visually see what agents have access to with connections and can search through them.
We are working on better organization and sorting now too to make it easier.
You can ask a single master agent for a status update. It will connect to all other masters and look at where things are. We've been thinking about even an ai-boss agent. Souren has a really cool protoype he can show you :)
@omri_ben_shoham1 @michael_yagudaev Adding to what Michael said. We're thinking rather than treating AI agents as "employees" that can "do stuff for you" instead think of them as AI supervisors that bring important things and decisions to your attention. A cross-canvas context awareness is a helpful step in that direction
Huge fan. I’ve been using it for the last month and a half - completely swapped out my use of an IDE. I went from Cursor, to Zed, now using canvas based Claude code/codex sessions in Agent Grid and really loving it.
As a product designer turned design engineer/builder - it’s really working for my spatial memory re: managing the various agent threads I’ve got running across various products. Maybe the best part? You can close it, and all your workspaces and the sessions/nodes are right where you left them. Can’t tell you how much anxiety I had of my IDE crashing and losing all the tabs/terminals I had open. 🤣
Excited to see where Michael and Souren take this.
@switmer777 You are a legend, Steve :). We might as well list you as a maker here with all the amazing feedback you have been giving us. Thank you so much! I cannot get back to building. There is soo much more to do 🚀
@switmer777 an OG, with utmost respect, having you as an early adopter has been amazing!!
The restart-anxiety story is very relatable, that's the actual reason most people give up on managing multiple agents. Curious about the depth of that persistence though - when you reopen the canvas, does it just restore the pane layout and terminal history, or does it actually resume an in-flight agent task that was mid-tool-call when the restart happened?
@galdayan Great question, I've seen it resume a session before. I think it depends on the harness and how it is implemented. So to give a better answer, I'll have to test it :). But in either case we should resume it.
One other user asked that we prevent the computer from shutting down if there is a worker that is working. He had macOS attempt to do a system restart on an overnight task, shut down AG and fail on another process (unrelated). So it failed both to apply patches and complete assigned agentic work.
So ideally, the computer would wait until it is free with an advisory warning, to let the harness know like "hey, we need to restart, wrap up and we will continue when we are back". Then on startup again just resume it.
That's what I'll try to add next week, hold me to it :)
@galdayan @michael_yagudaev This is a good point, I think partly we're moving towards more persistence beyond the tab/window restart. We've shipped daemon support (kinda like Docker) and we're experimenting with continued persistence in that fashion
@Souren Khetcho that supervisor framing is a genuinely different UX problem than the employee one though - a supervisor needs to know when to interrupt you versus when to just log something quietly for later. has that cross-canvas awareness actually changed how proactively the agents surface things, or is it still you pulling status by checking each canvas?
@omri_ben_shoham1 as it currently stands the status indicators is what I use to tell when a task is completed. Yes basically what we need to have is some form of a decision queue in a way or I guess like an agent that could surface these decisions as they happen. By nature it should just bring up things as they complete so that it can always keep the human in focus. I think there is a good UX approach where surfacing decisions can be non-obstructive.