Remix turns a plain-language product change into a live, isolated copy of the real app plus a ready-to-review pull request, aimed at the ticket-and-sprint-wait cycle non-engineers hit every time they want to test something on the live product.
Someone on the growth team notices the onboarding flow is losing people at step three. They know exactly what to change: shorten it, move the value prop up, kill a field. The fix is small. The queue to get it built is not. It goes into a backlog, gets triaged, and if it survives prioritization it ships whenever an engineer has room for it, weeks from now, if at all. Remix launched on Product Hunt today betting that the gap between "I know what should change" and "someone with commit access gets to it" is the actual product to sell.
Remix's pitch is specific enough to hold up. Connect a codebase once, and after that any invited teammate, a PM, a designer, a marketer, gets a live, isolated copy of the real app tied to its own URL. They describe the change in plain language inside Claude Code, Cursor, or whatever AI tool they already have open. What comes back is not a mockup. It is a running version of the actual product, safe to share in Slack, drop into a customer email, or point an ad at, alongside a "Remix request" that hands engineering the full prompt history, every file that changed, and a live preview link to click through before they approve, request changes, or merge. Nothing reaches production until an engineer signs off.
That review artifact is the part worth sitting with. A pull request generated by an AI agent is usually a wall of diffs with no story attached, which is exactly why engineers are slow to trust them. Remix's answer is to make the prompts themselves the reviewable object, so an engineer reading a request sees the reasoning that produced the code, not just the code.
What it costs
Remix's pricing page lists a free tier covering up to two collaborators, unlimited apps, forks, and branches during early access, and a one-time five-dollar cloud credit, then it stops. Pro runs $29 per seat per month for up to five collaborators with full cloud access. Teams is $99 per seat per month, unlimited collaborators, and adds role-based access control. Enterprise is a custom quote for SSO, SCIM, audit logs, and a self-hosted option, the tier a compliance team would actually need to say yes to.
Read the per-seat price as a floor, not a ceiling. The pricing page states plainly that the subscription and the cloud agent credit are billed separately, and every paid tier lists "cloud agent credit top-ups" as its own line item. The $29 or $99 buys a seat in the room. Running the agent against a real codebase, repeatedly, while a team iterates on a variant, is a second bill that scales with how much the tool actually gets used, the opposite of the flat, predictable cost the free tier implies.
There is no verified figure here for what the ticket-and-sprint version of this costs a company in engineering hours, so this piece will not invent one. What is verifiable is the shape of the trade: a team that used to wait on a sprint now waits on however long it takes to write a prompt and get a merge decision, and pays a per-seat subscription plus metered usage instead of a chunk of an engineer's calendar.
Who it is wrong for
This does not replace engineering, it puts a filter in front of it, so a team with no engineers to review the requests gets none of the value, just a sandbox full of unreviewed ideas. It also assumes the company already has people using Claude Code or Cursor day to day; if AI-assisted coding is not already part of how the engineering org works, Remix is asking a team to adopt two new habits at once, not one. A two-person pilot fits the free tier fine, but a five-person growth team that wants everyone trying variants is already paying Pro rates plus credit top-ups, and a regulated company that needs SSO or a self-hosted deployment is off the published price list entirely and into a sales call.
The premise also leans hard on engineering actually wanting the story, not just the diff. A team whose review culture is already "rubber-stamp whatever passes CI" will not suddenly slow down to read someone's five prompts because a tool handed them a transcript instead of a diff.
Product Hunt launches make a lot of "no more waiting on engineering" promises. The honest version of that promise still ends with an engineer's approval in the loop, which was always the actual constraint. Remix just moves the queue earlier, to before the idea gets typed, instead of after.