Configuration
AI build debugger
Let MarvX's AI diagnose failed deployments, review and apply its proposed fix, what it can change and see, and monthly AI analysis limits.
The AI build debugger reads a failed deployment's error and logs, explains the likely root cause, and proposes fixes MarvX applies only after you approve.
Use it
- Project → AI Debug: click Scan deployments. MarvX finds the latest failed deployment and shows an AI diagnosis (root cause, explanation, actions). Click Approve & redeploy to apply the actions and queue a deployment, or Reject.
- Intelligence page (Deployment Intelligence): failed deployments for the selected app, with Approve & apply or Reject.
- Project → Patches: AI fixes proposed for that app; apply or dismiss.
When a deployed container fails to start or keeps crashing, an analysis is queued automatically and appears under Patches. For build errors, run Scan deployments yourself.
What it can change
| Action | When approved |
|---|---|
| Set environment variable | Adds or updates one variable; others untouched. It never invents secret values; it tells you to add them. |
| Change build type | Nixpacks, Dockerfile, Static or Buildpacks |
| Add a config file | Commits a small file (Procfile, .nvmrc, .python-version, nixpacks.toml) to your GitHub branch ("Marvex AI: add to fix failed deployment") |
Adding files needs a GitHub repo and connection; otherwise add the file yourself. Some diagnoses only give instructions.
What it sees
Repository, branch and build type; environment variable names only (never values); the deployment error and log tail.
Limits
AI analyses per month: Free 3, Starter 20, Pro 100, Enterprise unlimited, resetting each calendar month. At the limit you'll see "monthly AI analysis limit reached". Paid plans can buy 25 AI fixes packs in Billing → Add-ons (credits never expire and are used after the allowance). Automatic analyses use only the monthly allowance, never credits.
MarvX Assistant can also explain a failed deployment, but it can't apply fixes.