NewInteractive Agents Live

Domain Experts view

Govern every run, from trace to outcome.

See exactly what your AI did, and why it did it. Then prove it to anyone who asks.

InteractiveAIProject A
Tomás Ferreira

Dashboard

9th Aug – 15th Sept
Past 24 hours
Environment: Default
Filters1

Traces

1,00050%

Sessions

20250%

Observations

20,07350%

Avg. Latency

4.34s50%

Avg. Session Duration

03.22s50%

Scores

39650%

Traces

Traces

Traces Tracked

1,000

1.6k1.4k1.2k1.0k0.8k
1/1/253/1/255/1/257/1/259/1/25

Report

CostRunsAvg. CostLatencyName
$0.002$0.002m 6s{name}
$0.002$0.002m 6s{name}
$0.002$0.002m 6s{name}
$0.002$0.002m 6s{name}
$0.002$0.002m 6s{name}

Models

All models
Cost by model

Cost

$0.07

5k3.75k2.5k1.25k0
1/1/253/1/255/1/257/1/259/1/25

Report

CostRunsCost/RunAvg. LatencyModels
$0.001.93M$0.07166268…6google/gemini-3.6-flash
$0.001.93M$0.07166268…4google/gemini-3.6-flash
$0.001.93M$0.07166268…2google/gemini-3.6-flash
$0.001.93M$0.07166268…2google/gemini-3.6-flash
$0.001.93M$0.07166268…2google/gemini-3.6-flash
support_session_842session replay
5 steps
Intent classified
Knowledge retrieved
Policy applied
PII redacted
Audit logged
Replayed end to end · on the record

Every session on the record · Tied to the policy that fired · Analytics across runs · Export the audit report

Every turn

Traced and replayable

Every decision

Tied to a named policy

Every edit

Named, reversible, no ticket

OpenTelemetry

Open, no lock-in

SOC 2 Type II ISO 27001 GDPR

Built to evidence EU AI Act Article 13 transparency and traceability obligations for high-risk AI, turn by turn. See Trust & Security →

Replay exactly what the AI did, in plain language.

Step through any session the way it happened: what the customer asked, which rule the agent followed, what it decided. No log lines, no engineering ticket, just the story of the run, ready to export as an audit report.

Step-by-step replay

Walk any run start to finish, in plain language.

Full conversations

Replay a customer's whole multi-turn session.

Plain-language view

Every step described in words, not log lines.

Every observation

Each step inside the run, captured as it happened.

Question

Analyze this refund dispute, ticket #8421.

Answer

See the operational pattern across thousands of runs.

Usage, containment, resolution, cost, and latency aggregate from the same record your auditor reads. The operational view of what production is doing, with nothing to reconcile and no second pipeline.

Usage and containment

Volume, containment, resolution, and transfer over time.

Cost and latency

Roll up across models and environments.

By use case

Compare operational health across teams and flows.

One source

The analytics read the same record the auditor does.

Across runslast 30 days
Auditable100%
P95 latency1.2s
Cost / session$0.04
sample data · built on the record

Prove which rule decided, and stand behind it.

Every answer traces back to the exact rule that produced it and the person who reviewed it. Roll any change back in one click, and hand the auditor a complete record. No engineering ticket, no waiting.

Which rule fired

Every answer linked to the policy that produced it.

Who changed what

Every edit logged with who, when, and what.

One-click rollback

Roll any change back to a previous version.

Auditor-ready

Generate the audit report on demand.

Audit report
Run#4f2a · kyc-verifier
Policy appliedkyc-enhanced-dd v3
Reviewed byMaria, 2 days ago
ResultPassed
Export PDFExport CSV
OpenAIAnthropicGoogleMistralDeepSeekMeta LlamaCohereMicrosoftand more

Bring any model

Any model you trust, run with governance and tracing.

Route any provider through one gateway with your own keys, every call attributed and on the record.

More on Governance

Read the thinking. See it shipped.

Browse all

Governance

Trust the AI you put in production.

Walk your security, compliance, and operations teams through sessions, traces, observations, analytics, and the audit report in one operational view. One session, end to end.