Let your people build apps with AI. Safely.
Security approves once what a type of app may do. The platform keeps every app inside those limits, acting with each user's own access, and records evidence for auditors on every release.
- Vetted building blocks only
- Security checks on every release
- Behind your company login
- No passwords or keys in apps
The problem
Your people already build apps with AI
Product managers, operations and analysts build the tools their teams need with AI assistants such as Claude Code, Codex or Cursor. You cannot stop it, and nobody can review AI-written code by hand.
Today
- Apps appear with nobody knowing what they do
- Passwords and API keys sit in the code
- Every release would need a code review nobody has time for
- No record of who changed what, or who allowed it
With aicoded.cloud
- Every app states what it may do, in plain language
- Apps hold no credentials at all
- Security approves app types once, not every release
- Every release leaves a change record for auditors
How it works
From a request to a running app
The builder talks to their AI assistant. The platform does the rest, and the assistant can trigger the delivery pipeline but never change or skip it.
Ask
An employee asks the AI assistant on their computer for an app.
Build
The assistant uses only the vetted building blocks of our open framework and previews the app locally.
Check
Our delivery pipeline builds it from the commit and runs the tests and the security checks.
Approve if needed
Within its app type it goes live by itself. More goes to a named approver, never the builder.
Run
The app runs where you choose, held to its declared limits at runtime.
What security approves
A permission list, not a code review
Security defines app types once: the audiences, data, email recipients, outside services and sizes an app of that type may have. Every app carries a permission list people can read in a minute.
- App type
- Internal tool
- Used by
- Hotel operations, Lisbon
- Reads
- Its own rooms and repair requests · SharePoint documents, with each user's own access
- Emails
- From ops@acme.example to company addresses only
- Outside services
- None
Ready on day one: Internal tool
The default app type, before security defines its own.
- Company staff only
- Company data only with each user's own access
- Email to company domains only
- No outside services
Every release lands in one of four levels
A preview with test data for a few invited people. No approval.
Within its app type, every check passed. Goes live without a human.
A new outside service, data or outside users. The approval card shows only what is new.
Anonymous public sites, payments, special categories of personal data, direct writes to core systems.
Guarantees
Enforced by the platform, not trusted to the app
App code can bypass a library, so every limit is held by the infrastructure around the app.
No credentials in apps
Apps reach their database, files, mail and secrets through the platform, never with passwords or keys of their own.
Each user's own access
An app reads company data with the permissions of the person using it, not with a shared account.
Private by default
Every app sits behind your company login. Who may use it comes only from its permission list.
Builder is never the approver
Checked by identity, with a fresh sign-in for every approval.
Only what passed runs
Every release is signed, and what runs is verified against the signature.
AI never touches production
The assistant's tokens cannot deploy, change data or approve. AI reviewers advise; deterministic checks decide.
Evidence for auditors
Every release leaves a record
A write-once change record for every release, sent to your security monitoring, plus an app register and a pilot report.
- Who asked
- The app's builder: “Add a photo to each repair request”
- What changed
- The code diff · permission list: + file store: photos
- Checks
- All security checks passed
- Who approved
- App type rule: Internal tool
- What went live
- The signed release, where it runs and when
App register
Each app's owner and users, the data it can reach, where it runs, its review date and its monthly infrastructure cost.
Pilot report
Apps live, time to launch, security findings and infrastructure cost per app.
Where apps run
Our hosting or your own cloud
The same app runs on either, unchanged.
Hosted by us
In the region you choose, with backups, monitoring and a data processing agreement.
In your own cloud
Your AWS, Google Cloud or Azure account. App data, logs and traces stay with you; we receive only permission lists, check results and change records.
Built on an open framework. Apps are built from the aicoded framework, open source under the Apache License 2.0, and can be built and run without the platform: you are never locked in.
aicoded.dev