What should your agent test?
Project / environment
Start by connecting your project
Run one command in your repository — AgentTwin maps the tools and workflows your agent uses.
Discovery
- 01Discover
Find tools and code paths.
- 02Review
Choose what needs testing.
- 03Run safely
Test through the CLI.
Selected resources
No project connected. Import the inventory from your CLI to see its dependencies and workflow candidates.
Environment map
0 resourcesEach card is a tool or service your agent uses — click one to see what its test needs.
Selected resources
No resources selected
Run init in your project, then import .agenttwin/inventory.json. You can also declare a resource manually.
Select software choose what your workflow uses
Runtime coverage
Node workflows with custom HTTP collections and bounded Slack operations run through the CLI. PostgreSQL and Redis worlds can use an explicitly selected cloud runner. Other provider adapters and browser/desktop execution require implementation. Native macOS apps need a compatible macOS runtime. Production accounts are not copied.
Project / workflows
Workflows
Pick something your agent does. Prepare a safe test, run it, and see exactly what passed.
Settings
Your account, projects and dashboard preferences.
Sign-in methods
Signed up with Google? Create a password to use email sign-in with the same workspace.
Project / runs
Test results
Every test run lands here — what was tested, what passed, and what to fix.
Execution queue
Connect your account in Settings to load submitted runs.
Loading evidence
Fetching recorded test results.