An app on Haechy can't reach anything on the internet that a person hasn't approved, and nobody can open it
without signing in. Below is each control, what it does, and where it stops.
Last updated . Every control is on for every app, on every plan; none of them is something you have to remember to switch on.
Outside services stay closed until a person approves them
An app has no route to the internet of its own. Its only way out is Haechy's outbound gateway, which lets a connection through only if the host is on that app's approved list. Everything else is refused and logged.
Claude can ask for a host, with a reason in plain words. Only the workspace owner or an admin can approve it, by opening the request in the Haechy console while signed in. Claude can't approve its own request.
An approval covers one host for one app. Your other apps stay as they are, and you can turn an approval off at any time.
Blocked attempts are recorded and Claude can read them, so it can explain what the app tried to reach and ask properly.
Each app has a Connections page in the console: every outside service it can reach, who approved it and the reason given, and the attempts that were blocked.
Package registries are open only while the app is being built; a running app can't reach them.
Expense tracker wants to connect to api.xero.com
Your app, on Haechy
Expense tracker
expenses-northwind.haechy.app
Outside service it would reach
api.xero.com
Secure connection (port 443)
Sign-off
Asked by
Claude6 Nov 202614:01 UTC
Held by
Haechy, until you decide
Your decision
Waiting for you
The top of an approval request, with the sign-off row. The box on the right stays empty until the owner or an admin decides.
A summary you can send to IT
Ask Claude for an app's security summary and it gives you a read-only link to pass on. It shows who can sign in, the app's region, its approved connections, when its last security check ran, its backups and the controls Haechy applies to every app. It contains no keys or passwords, and the person you send it to doesn't need a Haechy account.
Each app runs in its own sandbox
Apps run in containers under gVisor, which puts its own kernel layer between the app and the server it runs on. Each app also gets its own private network.
The file system is read-only, apart from a temporary folder.
The app runs as an ordinary user with every Linux capability removed and no way to gain more.
Memory, processing and the number of processes are capped to the plan.
The app never listens on the internet directly. Requests reach it only through Haechy's sign-in gateway.
Nobody opens an app without signing in
Haechy checks who someone is before their request reaches the app. A new app is open only to you and the people you add. You can widen it to everyone in your workspace. On paid plans you can also open it to anyone with an email address at your company's domain, or open the app to the public.
Before an app goes public, Haechy lists the pages and endpoints that anyone on the internet could then open without signing in, along with the outside services the app is approved to call. If it can't list them all, it says so. The app goes public only after you confirm that list in the Haechy console. Claude can't confirm it for you.
Colleagues sign in with a link sent to their email; they don't need a Haechy account. Haechy then tells the app who they are, so Claude never has to write a login system, which is one of the easiest things to get wrong.
Apps live on haechy.app, a separate domain from haechy.com, so an app can't read or set cookies for Haechy's own site.
We also went through every other way a request could reach an app without passing that sign-in. Apart from public mode, which is your choice, none of them lets someone who hasn't signed in reach the app's code. Scheduled jobs and the check after going live are started only from Haechy's own server. File links the app creates are signed and expire after 5 minutes unless the app asks for longer, and 1 hour at most. Outside services can't send webhooks into a private app. Each app has its own private network, and its container publishes no ports of its own.
Its own database, backed up every night
Each app gets its own Postgres database with its own login, which can connect to that database and no other. The database is backed up every night. Free keeps 7 days of backups; paid plans keep 30 days of nightly backups and 90 days of weekly copies (see plans). On every plan you can restore a backup by asking Claude; how restore works.
Every version of the app's code is kept as well. If a change goes wrong, ask Claude to go back to the previous one. That rolls back the code, not the data; a backup restore covers the data.
Keys and passwords for outside services are stored as app settings, outside the code, and handed to the app when it starts.
A check on every deploy
Every time Claude deploys, on every plan, Haechy runs an automated check on the code before it goes live. It looks for:
Keys or passwords written into the code, and uploaded .env files. These always stop the deploy until the values move into app settings.
Packages with known vulnerabilities or flagged as malicious. Haechy reads the exact versions from the app's lockfile and checks them against the OSV database.
Risky code: request input run as code, as a shell command, as SQL or in a template; secrets returned in a response; TLS checks turned off; open CORS; debug mode; admin or debug routes with no protection.
Outside services the code calls, listed with whether each one is approved yet.
A malicious package, a critical vulnerability in a package the app runs with, request input passed straight to code, a shell, SQL or a template, or a route that returns every secret also stops the deploy. The workspace owner or an admin can allow one of those exact findings from a link in the console; Claude can't. If the package version or the line changes, the check decides again. Everything else is shown as a warning, with the fix.
Example deploy checkDeploy stopped
A key is written into the codeA Xero client secret in src/xero.js, line 4. Move it to app settings and deploy again.
Outside service not approved yetapi.xero.com, mentioned in src/xero.js
No .env file in the deploy
It is an automated check for known vulnerabilities and the named risky patterns above. It doesn't find every bug, it can't see a key in a format it doesn't recognise, and it doesn't replace a careful look at what the app does.
Once the app is live, Haechy also opens its pages and passes any errors back to Claude with a hint on the fix, so Claude can sort them out before it tells you the app is ready. That check is about whether the app works, not about security.
Your app's data lives in the region you choose
Each app lives in one region. Its container, its database and its backups are all there. You choose the United States, the European Union, the United Kingdom or Australia.
Keeping the app's data in your region keeps it under that region's privacy law (GDPR in the EU, UK GDPR in the UK, the Privacy Act in Australia) and means no international transfer of that data to arrange. Every region runs on a US cloud provider, so this doesn't put the data out of reach of other countries' laws.
Some things are not tied to your app's region. Your Haechy account, sign-in and billing run on Haechy's control plane in the US. The full list of what lives where.
Some services we use to send email, verify phone numbers and take payments process data outside your chosen region. See our sub-processors.