Invokeable documentation
Connect your product, test customer journeys with AI agents, and fix reliability problems using verified evidence.
Invokeable tests how AI agents use your product. It checks the resulting product state, records the evidence behind each decision, and turns repeatable failures into findings your team can fix.
Get started
Connect a test environment and run your first assessment.
Understand Invokeable
Learn how catalog actions, agents, journeys, runs, evidence, and findings fit together.
Use the CLI
Run the same customer workflow from a terminal or automation.
Troubleshoot
Resolve setup, connection, assessment, and access problems.
The customer workflow
Connect a safe product environment
Add an MCP endpoint, an OpenAPI document, or a JSON file. Use a sandbox or staging system for actions that can change data.
Review the action catalog
Confirm what each action does, whether it can make changes, and whether Invokeable can verify its result.
Connect the agents you want to test
Invokeable checks compatibility before an agent becomes available for assessments.
Describe a customer journey
Write the task in ordinary customer language. Invokeable prepares the success checks, protected behavior, recovery expectations, and test cases.
Run, review, fix, and verify
Start the assessment, inspect the result and evidence, then verify any fix with a new run.
Choose a section
| Section | Use it when you need to |
|---|---|
| Concepts | Understand how Invokeable represents and verifies your product |
| Guides | Complete a task in the web application |
| Account | Manage the workspace, people, billing, and data |
| CLI | Work from a terminal or build automation |
| Reference | Check exact meanings, roles, statuses, and terms |
Start with Get started. If Invokeable is already configured, open Run an assessment.