What is AI pentesting
Aikido brings AI-powered penetration testing to Lovable. It lets you test your projects for real, exploitable vulnerabilities at any point in your project lifecycle. AI pentesting is automated penetration testing performed by AI agents that simulate real attackers. It performs dynamic analysis by interacting with your application to identify vulnerabilities that can be exploited. This includes sending real payloads, testing authentication and authorization flows, attempting privilege escalation, and probing APIs for unexpected behavior. This differs from Lovable’s built-in security scanner, which performs static analysis. The security scanner reads your code to flag known patterns such as exposed secrets, missing row-level security, and common misconfigurations. The two approaches are complementary.- Static analysis tells you what could go wrong based on your code.
- Dynamic analysis tells you what actually breaks when someone tries.
When to use AI pentesting
- Before a major release
Run a pentest after feature work is complete to catch vulnerabilities before going live. - After significant changes
Re-run after changes to authentication, database schema, or APIs to check for regressions. - For compliance or client requirements
Use the generated report to satisfy SOC 2, ISO 27001, or enterprise security questionnaires. - As an ongoing practice
Run regular pentests on active projects that handle real user data.
How Aikido tests your application
Aikido uses two testing approaches:- Blackbox testing
Attacks the application with no prior knowledge of the code, scanning to discover features, endpoints, and APIs - Whitebox testing
Uses full access to the source code to reason about application logic, roles, and data flows, catching logic flaws and access control issues that surface testing alone would miss
Prerequisites
- A paid Aikido account. Pentests are billed through Aikido.
Until the end of June 2026, each pentest costs 100 Aikido credits.
- Lovable workspace admin or owner role to connect Aikido.
- Lovable project editor or higher to run pentests and sync findings.
- A dedicated test user account in your Lovable project with username and password that Aikido can use to log in and test authenticated flows.
How to connect Aikido
A workspace admin or owner connects the Lovable workspace to Aikido via OAuth. Only one Aikido connection can be added per workspace. When a connection is created, it is shared with all workspace members and AI pentesting is available across all projects in the workspace.Add and name the connection
- Click Add connection.
- Enter a name for the connection, for example,
Aikido.
How to run pentests and fix findings
Start a pentest
Each pentest is scoped to a specific project.
- Open your project, then go to Security view → Agentic penetration test by Aikido.
- Click Prepare Aikido pentest (first time) or Launch new pentest (if you’ve run one before).
- Click Acknowledge and proceed to acknowledge the warning about database changes.
Review and complete setup in Aikido
Most configuration is already pre-filled. In most cases, you only need to:
- Add test users
Add at least one test user and provide the username and password in the Authentication instructions field. Click Save and Test to verify credentials. Without credentials, testing is limited to unauthenticated flows. - Review the remaining configuration in Aikido
Test scope, allowed domains, code & documentation, safety settings, and pricing are already set. You can adjust them if needed, but no changes are required to proceed. - Run the assessment
When preflight completes, click Run Assessment. The Confirm AI Pentest confirmation dialog appears. Read the checklist, tick the confirmation checkbox, and click Run Assessment to launch.
Sync findings to Lovable
The pentest runs for several hours, and issues appear in Aikido in real time as they are discovered. Findings are not synced automatically.When the pentest completes:
- Go to the Security view in your project.
- Click Sync findings to pull results in. Findings appear in the Agentic penetration test by Aikido section. Severity maps as follows:
- Critical and high → Error
- Medium → Warning
- Low → Info
Fix issues
- To fix an issue, copy the attack analysis from Aikido and send it in the chat, or click Try to fix all and Lovable will attempt the fix.
- To verify a fix, use the Retest issue option in Aikido, then sync again.
Get your pentest report
When the pentest completes, Aikido automatically generates a report combining an executive overview with actionable findings. The report is structured for SOC 2, ISO 27001, client security questionnaires, and investor due diligence. You can access and download it from Aikido.
Manage the Aikido connection
Workspace admins and owners can manage the Aikido connection from Settings → Connectors → Shared connectors → Aikido.- Reconnect: re-runs the OAuth flow to get updated credentials. A new window will open - make sure your browser allows pop-ups.
- Delete: permanently removes the workspace connection and its credentials. This cannot be undone. Workspace members will no longer be able to run AI pentests with Aikido. Existing synced findings remain visible in each project’s Security view.