Platform
Solutions
Customers
Resources
Pricing Book a demo
Platform · Audit trail

Every run, logged for governance.

Each workflow run is recorded end to end — completed and failed steps, per-step state, timestamps, inputs, and outputs. Human approvals are captured in the trail, so you can prove what happened and why.

Capabilities

What it does.

Enterprise-ready

Built for scale, governance, and reliability.

Composable

Works with the rest of the agentic platform out of the box.

Observable

Every run is logged, retrievable, and auditable via the Workflow Runs API.

Traceable

Per-step state and timestamps show exactly what ran, when, and how it ended.

Accountable

Human-in-the-Loop approvals are recorded as part of the run.

Exportable

Retrieve any run by ID; step outputs are downloadable where available.

The run record

Every run is a complete, retrievable record.

When a workflow runs, the platform logs the run with its inputs, overall state, and start and end timestamps. Retrieve any run by ID through the Workflow Runs API to monitor progress or reconstruct exactly what happened.

What gets logged

Steps, states, timestamps, and approvals.

Each run separates completed_steps from failed_steps. Every step carries its own state and timing, so nothing about an execution is left implicit.

state: DONE
state: FAILED
state: USER_ABORTED
started_at / ended_at
inputs
outputs
recorded approvals
Success and failure

Failures and aborts are first-class, not silent.

Completed steps record their output; failed or aborted steps record state and an error message. A user-intentional stop is marked USER_ABORTED so you can filter it apart from a genuine FAILED — clean signal for monitoring dashboards.

Why it matters

Governance you can hand to an auditor.

Prove what ran

Reconstruct any execution from its logged inputs, steps, and outputs.

Accountable approvals

Human-in-the-Loop checkpoints record who approved what before it proceeded.

Monitor at scale

Poll runs by ID to track progress and catch failures as they happen.

Distinguish intent

USER_ABORTED vs FAILED tells deliberate stops apart from real errors.

Example

A governed publishing pipeline, fully logged.

Input: draft brief AI Agent: research + draft Human in the Loop: approve copy LLM: format (JSON) CMS: publish Workflow Runs API: retrieve trail

The run logs each step's state and timestamps; the Human-in-the-Loop approval is recorded before publish, so the trail proves a person signed off.

How it works

From input to outcome.

01

Configure

Build it visually in Studio and add a Human-in-the-Loop checkpoint where sign-off matters.

02

Run

Trigger the workflow; it executes asynchronously while each step is logged with state and timestamps.

03

Approve

The run pauses at the checkpoint until a reviewer approves — and that approval is recorded.

04

Audit

Retrieve the run by ID to review completed and failed steps, inputs, and outputs.

Get started

Show us the workflow.
We'll show you the 10x.

Bring the marketing workflow that eats your week. We'll build it live, with your data and your models, in 30 minutes.