Skip to content
axintdocs

Cloud Check Loop

Axint and Cloud should not feel like two unrelated products.

The loop is:

  1. Build with Axint
  2. Read the check
  3. Fix what failed
  4. Rerun
  5. Ship when the result is clean

Axint is the local compiler + validator loop.

It gives you:

  • generated Swift
  • Apple-facing diagnostics
  • a lightweight Axint Check summary
  • a richer Fix Packet with the AI-ready repair brief

That is the fast local path.

Cloud is the saved-report layer on top of that loop.

Use Cloud when you want:

  • a free hosted check before you keep going
  • a saved report
  • a shareable result
  • a history of checks
  • an exact monthly allowance and account-visible usage meter
  • a Pro repair prompt you can give back to your AI

So the mental model is:

  • Axint runs the check
  • Cloud Check gives one anonymous verdict, then five signed-in checks each calendar month
  • Cloud Pro includes 50 checks each billing month, saved history, and the full repair prompt workflow
Terminal window
axint compile intents/LogWorkout.intent.ts --out ios/Intents/LogWorkoutIntent.swift

Axint emits:

  • .axint/fix/latest.check.json
  • .axint/fix/latest.check.md
  • .axint/fix/latest.json
  • .axint/fix/latest.md
Terminal window
axint xcode packet --kind validate
axint xcode packet --kind validate --format prompt

The first command gives you the latest result plus the top repair context.
The second gives you the full repair prompt.

Use Cloud when you want the result to become a report instead of staying only a local artifact.

That is where:

  • the result becomes shareable
  • the history becomes durable
  • the signed-in repair prompt becomes copyable
  • the loop is easier for teammates to follow

Feedback when Axint caused or missed something

Section titled “Feedback when Axint caused or missed something”

When Cloud Check finds a non-passing result, the report also includes a privacy-preserving learningSignal. That signal is not the source code. It is the small bug report Axint needs in order to improve:

  • stable fingerprint
  • likely owner
  • priority
  • diagnostic codes
  • source/output shape
  • suggested engineering action

Agents can print just that signal:

Terminal window
axint cloud check ios/Intents/LogWorkoutIntent.swift --feedback

Or through MCP:

axint.cloud.check({ sourcePath: "ios/Intents/LogWorkoutIntent.swift", format: "feedback" })

Use the repair prompt to fix the current project. Use the feedback signal to turn repeated Axint failures into fixtures, generator fixes, validator rules, or docs updates.

The right user experience is not:

  • read a wall of diagnostics
  • manually figure out what matters
  • copy random fragments around

The right user experience is:

  • read the latest packet
  • see pass / needs review / fail
  • copy one repair prompt
  • rerun

move from docs to workflow

Use the next step that matches what you need

If you just need a result, run Cloud Check. A free GitHub account includes five checks each calendar month with no card. Cloud Pro is $29 per month for 50 checks and the full repair workflow.