Check and fix a Lovable app with Scout

Connect Flamio to Lovable over MCP, let Lovable's agent run Flamio Scout on your published app, and have it fix what Scout finds.

Connect Flamio to Lovable once. Then Lovable's agent can run Flamio Scout on your published app, read each finding with its fix, and change the code — all from the project chat. Cursor, Claude Code and Codex work the same way.

Before you begin

You need the Editor, Admin or Owner role to create a token and start Scout.
Use a personal access token. A per-project key can read Scout runs, but cannot start one or mark a finding fixed.
Publish the app first. Scout checks a public address, such as https://your-app.lovable.app.
Each Scout run spends tokens from your balance: the walk stops at the run's budget, and assembling the report adds a little on top. See Settings > Usage.
You add the connector in Lovable. Lovable lets its agent use it in the project chat only, never in your published app.

Connect Flamio to Lovable

Create a token

In Flamio, click Settings, then API & MCP. Click Create token, enter a Name such as "Lovable", and click Create. In Your new token, click Copy.

Copy the MCP URL

Under Connect a client, copy the MCP URL: https://api.flamio.org/api/mcp.

Add Flamio in Lovable

In Lovable, open Connectors, click Add custom and select MCP server. Enter Server name "Flamio" and the MCP URL as Server URL. Select Bearer token or API key, paste the token, and click Add server.

Ask Lovable to check the app

In your project chat, enter:

Check my published app with Flamio Scout, then fix what it finds.

Lovable shows an approval card before the agent starts Scout, because a run spends tokens. Reading results needs no approval.

Result

Scout opens the published app in a real browser and walks it like a first-time visitor. A run takes minutes on a small app and longer on a big one. Ask Lovable again when it is done.

The agent then reads every finding: what confuses visitors and why, the pages and screen size it was seen on, a suggested fix, technical detail for the code, and a screenshot link. It changes the code, most severe findings first.

The run also appears in Flamio, under Flamio Scout, with the full report.

Check the fix

Publish the fixed version.
Ask the agent to mark the findings it fixed in Flamio and to run Scout again on the same address.
If the Flamio SDK records your app, open the run in Flamio Scout. A full day after the fix, it compares real visitors on those pages before and after.

What Scout does through Lovable

Step	What happens
Start a run	Read-only: Scout reads and clicks through the app, but never types, submits a form or signs in.
Same check twice	A run already under way on the same address, screen size and path is reused. Nothing new starts and nothing is spent twice.
A path it may not finish	A path that needs typing or submitting stops at that step, and the agent is told so — it is not a pass.
Follow a run	While Scout walks, the agent sees how many pages it has checked and the findings confirmed so far.
Read the findings	Usability findings, accessibility, technical errors, and look and feel, each with a fix.
Mark a finding fixed	The run page in Flamio shows the finding as fixed and compares real visitors before and after.

To test pages behind a login, or to let Scout fill in and submit forms, start an Interactive run in Flamio under Flamio Scout.

Check and fix a Lovable app with Scout

On this page