Run AI-powered penetration tests with Aikido
Aikido lets AI simulate real hacker attacks on your app to find vulnerabilities that actually break under pressure. Once the test finishes, you can import the results directly into Lovable and ask the AI chat to fix them.
Anyone preparing to launch an app, protect sensitive customer data, or meet compliance requirements like SOC 2.
Do this, in order
- 1
Create a dedicated test user account with a standard username and password inside your own Lovable app.
The testing AI needs login details to test screens and features that are hidden behind your login page.
- 2
Open Connectors from your workspace settings, select Aikido, and complete the connection popup.
This links your Lovable workspace to your paid Aikido account so projects can run tests.
- 3
Open your project's Security view, select Agentic penetration test by Aikido, and click Prepare Aikido pentest.
Lovable uploads your code repository and sets up the test environment in Aikido.
- 4
In the Aikido setup screen, enter your test user login credentials under Authentication instructions, click Save and Test, then click Run Assessment.
Aikido verifies that it can sign in and launches the automated attack simulation.
- 5
After the assessment finishes several hours later, return to your project's Security view and click Sync findings.
Results do not transfer automatically; syncing pulls the confirmed vulnerabilities into Lovable.
- 6
Copy the attack analysis from a finding into your project chat or click Try to fix all to apply code patches.
Lovable uses the exact steps the attacker used to rewrite your code and close the security hole.
Paste this into your project
Review our application authentication, API endpoints, and database permissions. Audit the project for unauthorized privilege escalation, insecure data access, and missing permission checks so we can pass our upcoming penetration test.
Words decoded
- Penetration testing (pentest)
- A simulated cyberattack where authorized tools try to break into your software to find weaknesses before criminals do.
- Dynamic analysis
- Testing security by actively sending inputs, fake logins, and malicious requests to your running app to see what breaks.
- Static analysis
- Scanning raw code files for common coding mistakes and leaked passwords without actually running the app.
- Whitebox testing
- A security audit where the attacker has full access to inspect your source code while planning attacks.
Where people get stuck
- If you skip adding test user credentials in Aikido, the test will only probe public pages and ignore your core authenticated features.
- Findings do not appear in Lovable automatically when the test finishes; you must manually click Sync findings in the Security view.
- Pop-up blockers in your browser can silently stop the Aikido authorization window from opening.
- Because the AI performs real attacks on your database and APIs, ensure you acknowledge potential test data creation before running.
The short version, steps, decoder and prompt on this page are written automatically from Lovable's own documentation and can lag or misread it. The official page is always the authority.