SDK
Application errors and user contextInstall the SDK for the runtime that owns the failure.
Verify: One real error arrives with environment, release, and stack context.
Production evidence for people and agentsSee what broke.
See what broke.
Know why.
Verify the fix.
Squasher connects errors, logs, traces, replays, deploys, and user impact in one cited incident record for engineers and coding agents.
Start with one real event. Use an SDK, a log drain, OTLP, or a guided agent setup.
INC-2814 · checkout-api
Illustrative incidentActive failure · 2m ago
Checkout failures after deploy
Impact412 usersError rate12.0%
Evidence attached
- Deploy 9c4a1f00:00
- Trace 8a71…b29c00:41
- Replay 7f13…38e100:52
Cited cause
Deploy 9c4a1f changed the currency parser. Null locale values now fail before payment authorization.
src/payments/currency.ts:48 · trace span 14
Fix verified
Tests passed · recovery held for 10m
One incident · connected evidence
One record from alert to recovery.
The evidence scope stays attached while Squasher moves from the first failure to a reviewed change and a recovery check.
- Detected
Failure crossed baseline
Checkout errors rose after deploy 9c4a1f.
- Correlated
Signals joined
Errors, logs, trace, replay, and release share one scope.
- Diagnosed
Cause cited
A changed parser and the failing span point to one file.
- Reviewed
Fix approved
A null guard and regression test are ready for review.
- Verified
Recovery held
Checks passed and the failure stayed below baseline.
One scope · every signal
Follow the evidence without starting the search again.
Move between product views with the same time, environment, service, trace, session, user, and query scope.
AskWhat broke after the latest deploy?
412 affected usersEvidenceSourceContext
TypeError in currency parsercheckout-api1,339 events
Payment authorization timed outpayments184 events
Cart request failedweb-checkout96 events
Why this view matters
Grouped failures with impact and ownership
Start from the failure that affects users, not a stream of repeated events.
Explore issuesAskShow logs in the failing trace before authorization.
38 matched recordsEvidenceSourceContext
locale resolved to nullcheckout-apitrace 8a71…b29c
currency parser rejected inputcheckout-apispan 14
payment request not sentpaymentsuser 8d2…71c
Why this view matters
Searchable records that keep incident scope
Time, environment, service, trace, and user context move with every drilldown.
Explore logsAskWhere did the checkout request begin to fail?
642 ms · 1 failed spanEvidenceSourceContext
POST /checkoutweb-checkout642 ms
parseCurrencycheckout-apifailed
authorizePaymentpaymentsnot called
Why this view matters
The request path and the slow or failed span
Follow one request across services without rebuilding filters or copying IDs.
Explore tracesAskWhat did the customer see when checkout failed?
00:48 session windowEvidenceSourceContext
Clicked Complete purchasecheckout00:31
POST /checkout returned 500network00:32
Error state renderedbrowser00:33
Why this view matters
The user session beside the technical failure
Keep clicks, network activity, console errors, and the related issue on one timeline.
Explore replaysAskDid checkout recover after the change?
10m recovery heldEvidenceSourceContext
Checkout APIUS West200 · 184 ms
Checkout APIUS East200 · 201 ms
Payment callbackEU200 · 226 ms
Why this view matters
Recovery checks tied to the incident
Use monitor results to verify the fix and keep watching for recurrence.
Explore uptimeAskDid the agent use the wrong tool or context?
1 failed tool callEvidenceSourceContext
Support agent runagenttrace 4c21…9a08
lookup_order tooltool callfailed
Customer feedbackresponsenegative
Why this view matters
Model and agent telemetry beside software signals
Inspect generations, tool calls, cost, feedback, and the services that handled the run.
Explore ai runsOne choice · one verification
Connect the path your stack already uses.
Choose one ingest path. Every path ends with the same proof: one real event in the correct project and region.
Log drain
Platform and deployment logsConnect the platform drain to the same Squasher project.
Verify: One deployment log appears in the live stream with its source.
OpenTelemetry
Existing logs, traces, and metrics pipelinesSend OTLP traffic to the collector for the project region.
Verify: One service and trace appear with the expected environment.
Prefer to configure it yourself?
Create a project, choose one recipe, and use Check again until the event arrives.
Trust and control
Fast answers still need a source and a safe next step.
Squasher is designed to help a responder move faster without hiding evidence, permissions, or the result of a change.
- Cited conclusions
- Cause summaries point back to source telemetry, trace spans, releases, and files.
- Reviewable changes
- Consequential actions stay proposed, approved, audited, and reversible.
- Region-aware ingest
- Projects keep their configured data region through ingest and query paths.
- One public contract
- The dashboard, API, CLI, MCP, and agent skills use the same product objects.
Pricing proof
Pay for production volume, not each collaborator.
Start on Free. Move to Pro when the production volume and workflow need it.
- Free
- $0/month
- Paid plans
- $29/month
Includes 10K error events each month.
Start with Pro.
AI triage is included on every plan.
View full pricing Start with one real event
Make the next incident easier to explain and safer to close.
Connect one service. Confirm one event. Keep the evidence scope through diagnosis and verification.
Connect your first service