Prepay a session
Each session costs €50 and opens one consecutive 60-minute API window.
Private beta · Metrics only
The Construct Metrics API measures AI agent and LLM workflow runs without taking control of your system. You run the workflow in your environment and send only structured observations.
No shell commands · No target URLs · No prompts, responses or credentials
How it works
Each session costs €50 and opens one consecutive 60-minute API window.
Call POST /sessions. The window begins when the session is created and cannot be paused.
Test your own agent or LLM workflow with your models, tools, RAG, memory and guardrails.
Submit status, timing, token counts, structured checks and error categories—never raw application content.
Get pass, error, timeout, latency, stability, tool-check and usage measures.
Safe contract
{
"run_id": "release-candidate-17",
"cost_currency": "EUR",
"cases": [
{
"case_id": "approval-01-run-1",
"status": "pass",
"latency_ms": 842,
"repeat_group": "approval-01",
"input_tokens": 122,
"output_tokens": 34,
"cost": 0.012,
"tool_checks": [
{
"name": "approval.read",
"passed": true
}
],
"boundary_violations": [],
"trace": {
"expected_steps": 4,
"observed_steps": 4
}
}
]
}{
"metrics": {
"schema_version": "1.0",
"run_id": "release-candidate-17",
"case_count": 1,
"status": {
"counts": {
"pass": 1,
"fail": 0,
"error": 0,
"timeout": 0
},
"pass_rate": 1
},
"latency_ms": {
"p50": 842,
"p95": 842,
"max": 842
},
"repeat_stability": {
"eligible_groups": 0,
"stable_groups": 0,
"stability_rate": null
},
"report_included": false
},
"idempotent_replay": false
}Private-beta price
Start the window with POST /sessions. Model and provider costs remain yours. VAT is added where applicable. Access keys are issued manually after a written B2B order and payment.
Request access
Share only non-confidential business information. We reply with fit, the private-beta terms and the next available access window.