Who Mushi Mushi is for

Mushi Mushi was built for one person first: someone who makes apps with AI tools and then has to deal with the bugs. Two other groups get just as much from it. Find yours below.

You build apps with AI, mostly on your own

You use Lovable, Bolt, Cursor or Claude Code. Your app has real users. You did not write most of the code by hand, and you don’t always know how it works.

What hurts

A user writes “it’s broken” and you have no idea where to start.
You ask the AI to fix it, the fix looks fine, and something else breaks.
You can’t watch your app all day, so you hear about problems late, or never: most people who hit a bug just leave.

What Mushi Mushi does

Your app gets a report button. Each report comes with a picture of the screen, the page and what the person tapped, and Mushi Mushi explains the likely cause in plain English.
Your AI editor gets a ready fix prompt. In Cursor or Claude Code you ask “get the latest Mushi report and fix it”; in Lovable or Bolt you paste the prompt.
It also looks for bugs nobody reports: a first-time user test, a watch on your live site, and a check for known security holes.

Start here: the install box on the home page (one command for Claude Code, one click for Cursor, a script tag for Lovable or Bolt). The free plan covers one app with 100 diagnoses a month, no card and no AI key.

You’re a developer and you own your code

You read the diffs. You might already run Sentry. You use AI agents, but you don’t merge what they write without checking it.

What hurts

Error tools tell you what the code threw, not what the person on the other end saw. Triage still eats your mornings.
Agent-written pull requests pile up, and each one needs a careful read.
CI is slow or expensive, and when it fails you can’t tell whether it was your code or the machine.

What Mushi Mushi does

Every report and every Sentry error lands in one list, with duplicates merged and a diagnosis that names the files and the steps to reproduce it.
Agents can open fix pull requests, and a second AI reviews each one against the original report before anything merges. When you choose “Merge when ready”, approved fixes land one at a time, each branch updated and its tests passed first.
Your CI bill and runners, explained: which workflows cost the most, what is wasted, and whether a failure was your code or the machine.
It works where you work: an MCP server for your editor, a CLI, and SDKs for web, React Native, Capacitor and Node.
It is open source (SDKs MIT, server AGPL), and you can run all of it yourself.

Start here: npx mushi-mushi in your app folder, then connect your editor from the connect page.

You run a team or a company

Several people ship. Bug reports come from support tickets, Slack, app store reviews and error monitoring, and nobody sees all of them.

What hurts

Bug reports are scattered, and the same bug gets reported five ways.
Tickets arrive without steps to reproduce, so engineers spend the first hour guessing.
You need to know where the data goes, who can see it, and what it costs.

What Mushi Mushi does

One list for every source: the in-app button, Sentry, low-star App Store and Google Play reviews, Slack, and a phone shortcut. Each report gets a diagnosis before anyone opens it.
Sends issues to Linear, Jira or GitHub Issues and alerts to Slack or Discord, so nobody has to leave the tools they already use.
Team seats and roles, an audit log, and a monthly spending cap you set.
For companies: single sign-on, longer data retention, and the option to run it on your own servers. See security and data residency.

Start here: compare the plans on pricing. The Studio plan includes 10 apps and 5 seats; Enterprise adds single sign-on and self-hosting support.

Not sure?

Try the sample diagnosis: one real-shaped bug from report to fix prompt, nothing to install. Or see everything else it does.