Give your customers visibility into every agent run

Embed GlassFlow's trace view into your product in one afternoon. Your customers see what their agents did, where they failed, and why, without ever leaving your platform.

app.acme.ai/observability
Alerts
Total Agent Runs149↓ -19.0%vs 184
Total Model Cost$1.53↓ -23.4%vs $2.10
Total Tokens534.0K↓ -16.4%vs 644.7K
Avg Score0.86No previous value
Total Agent RunsToken consumption by model
$1.53
Total cost
ModelInputOutputCache r.Cache w.
gemini-2.0-flash176.3K58.2K94.6K23.9K
gemini-2.0-pro134.7K81.5K42.9K11.8K
gpt-4.1182.1K29.5K73.9K15.7K
gpt-4.o-mini123.4K78.9K91.2K27.8K
Total Usage by AgentGrouped by agent
534.0K
Total tokens
TokensCost
Analyst Agent86.2K
Ops Agent64.1K
InputOutput
ScoresEvaluation scorer performance
avg 0.86
All scorers
Over TimeSummary
Completeness 0.93Status 0.84Actionability 0.93
Enterprise
Trace VolumeRuns and call count
4.5K
Total runs
AgentsWorkflowsTools
Triage Agent156
Forecast Agent138
LatencyHourly p50 and p95 latency
5052ms
Avg p50
AgentsWorkflowsTools
p50p95
Enterprise
Share traces with support
Upgrade to Pro for observability

See your agent runs, costs and traces.

Upgrade
Your customers
Customer A
Enterprise
Customer B
Pro
Customer C
Free

The problem it solves

Your customers are flying blind

They run agents that take minutes or hours and have no idea what happened inside them when something goes wrong.

You are paying the support cost

Every failed agent run that your customer cannot debug themselves becomes a support ticket you have to answer.

Building this yourself takes months

Trace collection, storage, query, and a UI that scales to your customer base is not a weekend project.

What your customers see

Acme
jordan@vantage.co
Your report took 3 minutes and 12 steps. Step 8 failed and was automatically retried.
Agent runDurationStatus
Daily digest generation1m 04sCompleted
Quarterly report3m 12sRetried
fetch-data
load-context
plan-report
gather-metrics
summarize
render-charts
step 8 · export-pdf
step 8 · export-pdf (retry)
deliver
Data sync0m 48sCompleted

This is GlassFlow. Your customers will never know.

A few of the views they get

Every run is a full trace. Your customers read a plain-language summary, drill into any step, and watch runs unfold live, all inside your product.

AcmeStep detail
Stepexport-pdf
StatusFailed, retried
Duration640 ms
Input{ format: "pdf", pages: 24 }
ErrorTimeoutError: render exceeded 30s
RetrySucceeded on attempt 2
AcmeLive run
Nightly data sync2m 18s elapsed
connect-source
extract
transform
load
verify

Manage every customer from one console

Spin up a workspace per customer and set their branding, retention, and what they can see. And yes, you can open any customer's traces yourself; each customer only ever sees their own, through their scoped embed.

Customers
VantageEnterprise
NorthwindGrowth
GlobexStarter
InitechGrowth
VantageEnterpriseView traces →
Workspacevantage
Product nameAcme
Primary color#14B8A6
Retention90d hot · unlimited cold
Embed tokenemb_live_9f2…a1expires in 7d
Visible columnsDurationStatusCost

Alerts, scoped to the customers who need them

Attach alert rules to a single customer or a whole plan. Enterprise accounts on an SLA get real-time paging; everyone else stays quiet, no noise and no extra cost.

EnterpriseVantage
Any run fails, or p95 latency > 2m
Slack #acme-slaPagerDuty
On
GrowthNorthwind
Error rate > 5% over 24h
Slack #alerts
On
StarterAll other customers
No SLA alerts
Off

What you control

Branding

  • Your logo and primary color
  • Your product name throughout
  • Sanitised error messages or raw technical detail, your choice
  • Which columns your customers can see (hide cost if you prefer)

Data isolation

  • Every embed token is scoped to one end user
  • Tokens expire on a schedule you set
  • One customer can never see another's traces
  • All data stays within your GlassFlow workspace
Email address copied