Case studyArchitect caught 15x more agent failures.19 hours a month back

Fix the agent failures that cost you customers.

Find every run where your agent went wrong, prove each fix on your own traces,
and monitor production so every fix stays fixed.

10M+agent traces monitored
84,000+silent failures caught
3,480+failure patterns found
how it works

From a silent failure
to a verified fix.

1/5Trace

Bring the traces you already have from Langfuse, LangSmith or OpenTelemetry. Every step and tool call in one place.

trc_01M5HDJWD0Q7RXK2T9V3BNCF8AOpen in Langfuse

Oct 22 9:03 am

TranscriptGraphOpen in Langfuse

Summary: Customer asks why they keep getting logged out on other devices; agent explains the session policy and suggests signing in again.

User

Why do I keep getting logged out on my other devices?

routerOK

Routed to → account_support

search_kbOK
InputQuery: logout other devices
OutputDocuments: 0
support-agentOK

For security, every session expires after 24 hours on all of your devices. Signing in again will sort it out.

send_replyOK
InputChannel: web chat
OutputDelivered: true

Deviations

A deviation is a recurring way your agents stray from their intended behaviour. Glassray watches your synced traces and surfaces new ones automatically.

Run discoveryReload
Add filter
Confirmed · 3
Refund issued before the order was confirmed
Process a refundbreaksConfirm the order before issuing a refund6 examples
Quoted the monthly price on an annual plan
Answer a billing questionbreaksQuote the price of the customer's own plan9 examples
Promised a callback but opened no ticket
Escalate to a personbreaksEvery escalation opens a ticket3 examples
Suspected · 1
Fabricated policy when the knowledge base has no answer
Answer an account questionbreaksAnswer only from retrieved help articles41 examples
Done · 4
agent-alerts
GlassrayAPP9:04 AM

Fabricated policy when the knowledge base has no answer · major

Suspected

When search_kb finds nothing, the agent answers anyway and states an account policy as fact.

Answer an account question

41 instances in 1,280 traces scanned · 41 major · last seen 41s ago

ConfirmDismissView in Glassray

3 repliesLast reply just now

Deviations

Fabricated policy when the knowledge base has no answer

Expected

When the knowledge base has no answer, say so and offer a handoff to a person.

Observed

After search_kbreturns no documents, the agent states an account policy as fact instead of admitting it doesn't know.

Trigger rate
3.2%41 of 1,280 evaluated traces triggered this
41/1,280
41 major

Measured across all 1,280 traces evaluated against this rule.

Trigger rate over time24h7d30dAll
Oct 9Oct 15Oct 22

41 examples over Oct 9 - Oct 22, against the 1,280 traces scanned in this flow.

trc_01M5…F8A41s agoView trace

search_kb found nothing for "logout other devices", and the agent told the customer every session expires after 24 hours.

Why the agent invents account policies

Why does the support agent make up account policies?

Deviation · Fabricated policy when the knowledge base has no answer
Worked for 14 seconds
Loaded the 41 traces in this deviationCompared them with 200 runs that answered correctlyRead the support agent's system prompt

All 41 runs start with an empty retrieval. search_kb returns no documents, and the system prompt tells the agent to always give a definitive answer, so it fills the gap with a policy that sounds right.

Suggested fix: when retrieval comes back empty, say so and offer a handoff to a person.

Deviation Fabricated policy when the knowledge base has no answer
Find traces, or paste a screenshot…
acme / support-agent

Hand off when the knowledge base has no answer #218

Opensam wants to merge 1 commit into main from fix/kb-empty-handoff
Conversation2
glassraybotcommented 2 minutes ago

Glassray simulation: 1 regression

Replayed 48 production traces from Fabricated policy when the knowledge base has no answer against fix/kb-empty-handoff.

  • Logged out on my other devicesmainInvents a policyThis PROffers a handoff
  • How do I reset my password?mainAnswers from the docsThis PRHands off instead
Same requestmainThis PR
Logged out on my other devicesInvents a policyOffers a handoff
How do I reset my password?Answers from the docsHands off instead

41 fixed · 1 regression · 6 unchanged · View the replay in Glassray

Deviations

Fabricated policy when the knowledge base has no answer

Done - holding across recent traces· Fixed by #218View PR
Trigger rate
0.2%3 of 1,412 evaluated traces triggered this
3/1,412
3 major

Measured across the 1,412 traces evaluated since #218 merged.

Trigger rate over time24h7d30dAll
Sep 23Oct 3Oct 13Oct 22

3 examples since the merge, against 1,412 traces scanned in this flow.

Activity
a fix landed · #218 mergedOct 13, 16:40
moved this to Done · holding across 1,412 tracesOct 20, 09:12
Read the docs
customer example · Super44

The briefing read perfectly.
The forecast was invented.

Super44 briefs local merchants on their own business every morning: an agent running up to 20 tools compiles sales, staffing, reviews, and the weather that feeds the revenue estimate. One morning the weather tool returned empty, and the agent wrote the briefing anyway, reusing yesterday's forecast as if it were confirmed.

No error, no complaint, nothing for an eval to fail. Glassray compared the reply against the tool outputs in the trace, found a forecast with no data behind it, and flagged the run with the evidence attached. The pattern stays monitored, so a made-up number never gets to become a habit the merchant plans around.

“Do merchants trust that when the agent tells them something about their data, it's true? If it says ‘you've sold more Aperol Spritzes’ and they think, I don't believe you, our value diminishes.”
Henry George
Henry GeorgeCo-Founder at Super44
Read the case study
Super44app.super44.ai
Tuesday briefing

Good morning! Today's sunny and mild, so expect the terrace to fill up. Projected revenue: €18,400.

Ask anything about your business...
What the merchant sees
Trace viewmorning briefing
  • ok
    Point-of-sale lookupsales data returned
  • ok
    Reviews + staffingreturned normally
  • problem
    Weather toolreturned empty, no data for today
  • problem
    Composerwrote briefing using yesterday's forecast
  • ok
    Final output‘today’s sunny and mild’ · fluent, confident
What Glassray sees
More examples
“
The statefulness of Glassray is the biggest win: it helps developer experience, reduces the noise, and helps us figure out what's important over time.
Henry George
Henry GeorgeCo-Founder at Super44
Read the case study
pricing

Observability is generously free.
Pay only when we catch mistakes your agent made.

We follow a "price should be a no-brainer" strategy. If it's not, tell us why.

Traces & observability
$0free tier

Every trace, in one place. Plug in your stack and see what your agents are doing.

  • SDK (native), OpenTelemetry, Langfuse, LangSmith, PostHog, Braintrust & these integrations
  • 150k traces / month, then $1 per 10k
  • 1-year retention
  • 1 project
  • Trace explorer & search
  • MCP, API & CLI access
  • Unlimited seats
  • A better UI than the others
Agent monitoring
$499/ month

Glassray watches production for you and flags the runs that quietly went wrong.

  • Everything in free
  • Silent-failure detection on every trace
  • Built-in failure modes + custom signals
  • 100 active confirmed deviations, then $5 each
  • Suggested code-level fixes
  • Destinations: push deviations to Slack / webhooks
  • 5 projects
Enterprise
Custom

A custom solution for teams running agents at scale.

  • Everything in Agent monitoring
  • Model post-training on your traces
  • Custom RL environments
  • SSO/SAML login
  • Audit logs
  • Dedicated customer success manager
pricing calculatorEstimate your cost
Alpha

Make your agents
self-improve.

Limited alpha spots. Join the waitlist and we'll reach out when your stack is ready to plug in.