Flagship dataset / agent action benchmarks

Ship agents with data you can replay.

A privacy-safe synthetic corpus for evaluating tool choice, checkout behavior, latency, retries, and recovery before production. Start with a free preview, then scale through API delivery or monthly refreshes.

Get free preview Start $49/month refresh Request $5k pilot Inspect schema

99.1%

replay consistency

Deterministic schemas for agent evaluation and regression testing.

10k+

benchmark rows

A ready-to-run sample pack for tool choice, checkout, and recovery traces.

Base USDC

agent-native delivery

Pay per row through x402 or choose a monthly refresh plan.

Free operator sample

For operators / watch first

See the render before you buy.

Watch the free sample lane, then open the video studio to build a one-minute physical sample with synchronized metadata and delivery options.

Open sample data Open one-minute studio

Operator sample library

View every sample lane before buying.

Open full studio →

Studio sample

Open the studio to generate this playable video lane.

Open video studio

One-minute studio sample

Open the video studio and generate a 60-second operator-ready physical sample with frame-level ground truth.

Studio sample

Open the studio to generate this playable video lane.

Open video studio

Enterprise UI

Generate and inspect synchronized interface sequences in the studio.

Studio sample

Open the studio to generate this playable video lane.

Open video studio

Industrial dynamics

Generate constrained motion and ground-truth packages for operators.

Dataset sample

Agent action traces

Tool choice, retries, checkout, latency, and recovery fields.

View sample rows

Dataset sample

Operator telemetry

Structured events for incident review, escalation, and SLO testing.

View sample rows

Dataset sample

Vision edge cases

Synthetic VFX and industrial scenarios with reproducible metadata.

View sample rows

What is included

A practical evaluation layer for autonomous systems

XITool-selection traces with success and failure outcomes
XICheckout and payment-flow scenarios with bounded retries
XILatency, recovery, and escalation fields for SLO testing
XIStable provenance metadata and synthetic disclosure on every row
XIFree preview plus API, bulk, and monthly refresh delivery

Machine contract

Built for agents, not spreadsheets

View plans →

GET /api/v1/data/agent-action-benchmarks/preview
GET /api/x402/v2/datasets/detail?id=agent-action-benchmarks
POST /api/x402/v2/answers