Over the last year my job moved from writing code to reviewing code that agents write. Running Claude Code, Codex and OpenCodeacross several projects, I kept losing track of what I had asked where and which tasks were actually finished. So I built VibeiDE for my own daily work.
Why it's different: most tools help you start more agents. But starting agents is the easy part. The bottleneck in software development has shifted from writing code to validating it, and that is what VibeiDE is built for.
How I work with it: I click through my app, and every time I spot something, I launch a new task in a sentence and keep going. Each task gets its own terminal and a fresh CLI session, started automatically. I don't have to remember anything, because when a task finishes it waits for my review, and then I either mark it done or send a follow-up. No more sitting and waiting on one or two terminals, and no more guessing which tab holds what. It's a ridiculously simple idea, and ridiculously effective.
What's inside:
- Project queues: you set the max parallel tasks per provider, and a task starts as soon as its provider has a free slot.
- Fresh session per task: every task keeps its session, so you can resume it or follow up any time.
- Review before done: a finished task waits for you. Mark it done, or send a follow-up in the same session.
- Git your way: run tasks on one branch, or give each task its own worktree.
- Terminal or chat: watch a task in its original CLI terminal, or switch to a clean chat view of the same session.
- Plan council: one agent drafts a plan, another reviews it before any code is written.
- Grid or Office: watch your agents as a grid of terminals, or in a 3D studio with a desk per agent.
- Private by design: end-to-end encrypted phone control, and voice input that runs locally on your computer.
In the last 30 days I spent 386 hours in VibeiDE, ran up to 14 agents in parallel, and got my first paying customer before this launch.
It uses your own CLIs and provider accounts and does not bypass provider limits. Runs on macOS and Windows, Linux is experimental. Free 14-day trial, no card. Launch prices ($7.99/month, $79/year or $249 lifetime) last until January 1.
In development:
- Company Mode 2.0: tackles the next bottleneck, several people generating code fast in the same project.
- Voice agent: full voice control of VibeiDE.
- MCP server: agents can create new tasks and control VibeiDE themselves.
- Built-in browser: so you never have to leave the app.
- Agency mode: a marketing harness, so solo builders can keep up with marketing too.
I put a lot of myself into this, so I'd love your honest feedback: what do you like, what don't you, and what's still missing in your multi-agent setup?
On every agent run I ask it at the end to list the changes it made that nobody asked for, it takes ten seconds and the list is never empty. Could the review screen show that part on its own, everything it touched outside the task?
The human Complete button is a good boundary, but it needs to expire. If the branch changes after tests or review, the task should return to Needs review automatically. I’d also surface the exact verifier command and exit code beside the diff, so a reviewer can separate ‘agent stopped’ from ‘acceptance check passed.’
Framing the bottleneck as validation rather than starting more agents rings true. Once several agents are running, the hard part is keeping track of what was asked where, and whether each result is actually finished. Making a finished task wait for review, with the same session available for a follow-up, is a sensible default — especially for changes touching auth, migrations, or database policies, where a silent auto-complete would be risky. Letting each task use its own worktree, while still allowing a single branch when that fits, also keeps the review step clean instead of tangling parallel work together.
Works great! :) Keeps things simple and manageable. Leveling and achievements are a nice touch! :D
When parallel tasks finish, how does the review screen show the diff, per task or combined across projects? Is there a way to mark a result as needing a rerun with new instructions?
A finished task waiting for review feels like the right default. With several agents running, there needs to be a clear moment where a person decides the change is actually done.
About VibeiDE on Product Hunt
“Queue agent tasks in parallel and review each result”
VibeiDE launched on Product Hunt on October 26th, 2026 and earned 85 upvotes and 15 comments, placing #18 on the daily leaderboard. Queue Codex and Claude Code tasks across your projects, run them in parallel and review every result. Built for developers and solo founders.
VibeiDE was featured in Productivity (664k followers), Developer Tools (521.9k followers) and Artificial Intelligence (482.1k followers) on Product Hunt. Together, these topics include over 394.8k products, making this a competitive space to launch in.
Who hunted VibeiDE?
VibeiDE was hunted by Patrik Kelemen. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how VibeiDE stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Patrik Kelemen
Hi Product Hunt, I'm Patrik 👋
Over the last year my job moved from writing code to reviewing code that agents write. Running Claude Code, Codex and OpenCode across several projects, I kept losing track of what I had asked where and which tasks were actually finished. So I built VibeiDE for my own daily work.
Why it's different: most tools help you start more agents. But starting agents is the easy part. The bottleneck in software development has shifted from writing code to validating it, and that is what VibeiDE is built for.
How I work with it: I click through my app, and every time I spot something, I launch a new task in a sentence and keep going. Each task gets its own terminal and a fresh CLI session, started automatically. I don't have to remember anything, because when a task finishes it waits for my review, and then I either mark it done or send a follow-up. No more sitting and waiting on one or two terminals, and no more guessing which tab holds what. It's a ridiculously simple idea, and ridiculously effective.
What's inside:
- Project queues: you set the max parallel tasks per provider, and a task starts as soon as its provider has a free slot.
- Fresh session per task: every task keeps its session, so you can resume it or follow up any time.
- Review before done: a finished task waits for you. Mark it done, or send a follow-up in the same session.
- Git your way: run tasks on one branch, or give each task its own worktree.
- Terminal or chat: watch a task in its original CLI terminal, or switch to a clean chat view of the same session.
- Plan council: one agent drafts a plan, another reviews it before any code is written.
- Grid or Office: watch your agents as a grid of terminals, or in a 3D studio with a desk per agent.
- Private by design: end-to-end encrypted phone control, and voice input that runs locally on your computer.
In the last 30 days I spent 386 hours in VibeiDE, ran up to 14 agents in parallel, and got my first paying customer before this launch.
It uses your own CLIs and provider accounts and does not bypass provider limits. Runs on macOS and Windows, Linux is experimental. Free 14-day trial, no card. Launch prices ($7.99/month, $79/year or $249 lifetime) last until January 1.
In development:
- Company Mode 2.0: tackles the next bottleneck, several people generating code fast in the same project.
- Voice agent: full voice control of VibeiDE.
- MCP server: agents can create new tasks and control VibeiDE themselves.
- Built-in browser: so you never have to leave the app.
- Agency mode: a marketing harness, so solo builders can keep up with marketing too.
I put a lot of myself into this, so I'd love your honest feedback:
what do you like, what don't you, and what's still missing in your multi-agent setup?