AI-Assisted App Management

The My Apps section includes a built-in AI assistant that lets you deploy, configure, and troubleshoot applications using natural language. Instead of navigating menus and filling out forms, you describe what you want and the assistant handles the steps.

What It Is

The AI assistant in My Apps is powered by the OSC MCP integration and understands the full lifecycle of a custom application — from initial deployment through configuration, monitoring, and debugging. It has access to your apps, parameter stores, databases, domains, and logs.

You interact with it through a prompt bar at the top of the My Apps page. Suggestion chips appear below the prompt bar with common actions you can take in one click. When one of your apps has a problem, a health banner appears at the top of the page to alert you and offer diagnostic options.

What the AI Can Do

Deploy Apps

Deploy an application directly from a git repository:

Deploy my app from https://github.com/youruser/yourrepo

The assistant will ask for the app name and runtime type (Node.js, Python, or WebAssembly) if you have not specified them. If the repository is private, it will ask for an access token. It will also warn you about common issues before deploying:

  • If your app reads PORT, it must use 8080 — the assistant will remind you of this if it detects a Node.js app
  • If your Next.js app uses NEXT_PUBLIC_* variables, those must be committed to the repo in .env.production rather than supplied via the parameter store, because they are baked in at build time

Set Up Configuration

Create a parameter store and bind it to your app in a single request:

Set up a parameter store for myapp with DATABASE_URL and SMTP_HOST

The assistant will provision a Valkey instance, create an App Config Service, add the keys you specified, and bind the parameter store to your app. You can also change or remove a parameter store binding on an existing app:

Bind the parameter store called myconfig to my app called myapp

Remove the parameter store binding from myapp

See Parameter Store for more on managing configuration values.

Diagnose Failing Apps

When an app shows a "Build failed" status, ask the assistant to investigate:

Diagnose why myapp is failing

The assistant will fetch recent logs, identify the likely cause (missing dependency, build error, wrong PORT value, etc.), and suggest a fix. You can also ask it to filter logs directly:

Show me the last 50 lines of logs for myapp

Show me errors in the logs for myapp since 2026-03-01T00:00:00Z

Manage Custom Domains

Map a custom or platform-managed domain to an app:

Set up a custom domain api.example.com for myapp

Give myapp a managed subdomain

For custom domains, the assistant will walk you through the required DNS configuration. See Custom Domains for the full guide.

Create and Connect Databases

Create a database and wire it to your app in a single conversation:

Create a PostgreSQL database for myapp and add the connection URL to its config

The assistant will create the database instance, retrieve the connection URL, and either create or update a parameter store with the DATABASE_URL value, then bind it to your app.

Back Up and Restore Databases

Trigger a manual backup or restore a database instance from a previous snapshot:

Back up my PostgreSQL database called mydb

List backups for my CouchDB instance called nosql

Restore my database called mydb from the most recent backup

The assistant uses the create-backup, list-backups, and restore-backup MCP tools. Backups require a paid plan. See Database Backups for the full guide.

Restart and Rebuild

Restart your app to pick up new commits, or rebuild to clear caches:

Restart myapp

Rebuild myapp

See Managing Custom Apps for when to use restart versus rebuild.

Example Conversations

Deploying a new app with a database

Deploy my app from https://github.com/youruser/myapi as a Node.js app called myapi,
then create a PostgreSQL database and add the connection URL to its config

The assistant will: 1. Deploy the app from the git repository 2. Create a PostgreSQL service instance 3. Retrieve the connection URL 4. Create a parameter store with DATABASE_URL 5. Bind the parameter store to the app and restart it

Diagnosing a build failure

My app called frontend is showing "Build failed". What went wrong?

The assistant will fetch the build logs, explain the error, and suggest what to fix. Common causes it can identify include missing package.json scripts, incorrect PORT usage, failed npm install steps, and syntax errors.

Switching to a new parameter store

I created a new parameter store called prodconfig.
Switch myapp to use that one instead of devconfig.

The assistant will update the parameter store binding on myapp and trigger a rolling restart so the new configuration takes effect without downtime.

Tips for Effective Prompts

Be specific about app names. The assistant manages all apps in your team, so saying "my app" without a name may lead to a clarifying question. Use the exact name you gave the app when deploying it.

Mention the runtime if it is not obvious. If you say "deploy from https://github.com/youruser/myrepo", the assistant will ask whether it is Node.js, Python, or WebAssembly. Include it upfront to skip that step: "deploy as a Node.js app".

Ask the assistant to confirm before destructive actions. For operations like rebuild or deleting a parameter store binding, you can ask it to describe what it is about to do before proceeding: "Tell me what steps you will take before doing anything."

Chain related tasks together. The assistant handles multi-step workflows in a single request. You do not need to issue separate commands for "create database", "get connection URL", and "add to config" — describe the outcome you want and it will work through the steps.

Use the health banner. When an app has a problem, the health banner at the top of My Apps provides a one-click path to start a diagnostic conversation with the context of your specific app already loaded.