Skip to main content
Synthetic tests run a simulated autonomous-agent checkout against a merchant and report exactly where the agent would succeed or fail: discovery, selection, cart add, checkout form, payment. Useful for quantifying integration gaps before investing engineering time.

​
POST /v1/scanner/tests

Scope required: tests   Cost: 5 credits
Request fields: Test types run a subset of steps — use browse for a quick check, full_flow for a complete checkout simulation. Response is an AgentShoppingTestResult with per-step status and a blockers array flagging specific friction points (missing structured data, CAPTCHA, account requirement, unsupported payment method, etc.).

​
GET /v1/scanner/tests/:domain

Retrieve previous test results for a domain. Scope required: read   Cost: 0
Returns the 5 most recent tests in descending tested_at order.

​
Interpreting results

A test with status: "passed" and success_rate: 100 means the agent completed every step. More common is status: "partial" or "failed" with populated blockers — these are the actionable findings. Blocker severity:
  • blocking — the agent cannot continue. Common examples: no_structured_data, captcha_required, account_required, javascript_wall.
  • warning — the agent continued but with degraded confidence (e.g., inferred price from OCR instead of a schema.org price field).
failure_point on the result root tells you exactly where the run stopped:

​
Economics metadata

When the merchant has category data, the test enriches the result with:
These are heuristic projections based on merchant-category baselines (visits, AOV, conversion rate) multiplied by the test’s failure rate. Use as directional signal, not as revenue forecasting.