Hybrid DaaS marketplace · v2

The Continuous Data Layer for Autonomous Digital Agents and Physical AI

Scale8.ai is the world's first hybrid data exchange delivering privacy-scrubbed execution logs, annotated static assets, and low-latency robotics telemetry streams directly to your models.

Explore Dataset Catalog View API Docs

8.4ms: p50 stream
1.2B: traces indexed
100%: PII scrubbed

{
  "trace_id": "trc_8f21c4a0",
  "model": "computer-use-preview",
  "steps": [
    { "t": 0.00, "action": "dom.click",
      "selector": "#invoice-export",
      "screenshot": "s3://scale8/redacted/0001.webp" },
    { "t": 0.42, "action": "tool.call",
      "name": "billing.export",
      "args": { "range": "2026-Q2", "email": "[REDACTED_PII]" } },
    { "t": 1.08, "action": "human.correction",
      "note": "chose CSV over PDF" }
  ],
  "reward": { "task_success": 1.0, "steps_optimal": 0.86 }
}

live sample · redacted · region eu-west-1

The infinite data loop

From Generation to Action

AI is moving from generation to action. But for an AI Agent to navigate software or a Robot to move through the physical world, it needs continuous context. Scale8 is the real-time data engine that bridges models to reality.

Just as an 8-note scale transforms raw frequencies into harmony, Scale8 structures unstructured action streams—scrubbing PII, standardizing schemas, and delivering low-latency trace feeds. We provide the infinite data loop (∞) that powers autonomous execution.

Dataset catalog

Three exchanges, one contract, one billing surface — queryable from a single API key.

AI Agent Traces

DOM interaction logs, API tool calls, and human decision trajectories for fine-tuning computer-use models.

DOM events
Tool calls
Trajectories

Static Asset Bundles

Structured workflow databases, SaaS UI state screenshots, and licensed code dumps.

Workflow DBs
UI states
Code dumps

Physical AI Telemetry

Zero-egress spatial video, LiDAR streams, and ROS2 manipulation logs from live fleets.

LiDAR
Spatial video
ROS2

Built for compliance-grade throughput

Automated Privacy Pipeline

Inline text and image redaction for PII across every frame, payload, and log line before it ever leaves the source.

Low-Latency Streams

Sub-10ms delivery of execution logs over persistent gRPC channels, with backpressure-aware replay.

Metadata & Rewards

Trace data arrives pre-validated with success metrics, step-optimality scores, and provenance signatures.

One key. Every exchange.

REST for catalog queries, gRPC for streams, S3-compatible signed URLs for bundles. SDKs for Python, TypeScript, and ROS2.

Get an API key

from scale8 import Client

c = Client(api_key="s8_live_...")
for step in c.traces.stream(
    domain="computer-use",
    min_reward=0.8,
    redact="strict",
):
    train(step)

Join the Waitlist

Tell us what you're training. We onboard partners in weekly cohorts.