Ask about funnels, errors, users, APIs, feature flags, replays, and more — and get answers grounded in your own product data.
Most analytics tools give you dashboards.
Dashboards are useful, but they still require you to know where to look.
If conversion suddenly drops, you might open your funnel. Then check errors. Then look at API performance. Then inspect session replays. Then wonder whether a recent feature rollout affected the same users.
The data may all be there. The hard part is connecting it.
That is what we built Ask TraceIQ for.
Ask the question first
Ask TraceIQ lets you investigate your product using natural language.
You can ask things like:
Why did checkout conversion drop today?
What errors are affecting the most users?
Are users exposed to this feature flag behaving differently?
Which API endpoints are currently failing?
Show me what happened before users abandoned this funnel.
Are our webhooks healthy?
TraceIQ can investigate across the product instead of forcing you to manually move between different dashboards.
It can inspect funnels, event trends, errors, API performance, feature flags, user timelines, session replays, retention, cohorts, alerts, and webhook health.
The result is one coherent investigation with the relevant evidence attached.
Not just another chatbot
We didn't want Ask TraceIQ to be a generic chat box sitting beside your analytics.
The AI does not receive unrestricted database access, and it does not generate arbitrary SQL.
Instead, it works through a curated set of read-only TraceIQ tools.
When you ask a question, Ask TraceIQ decides what information it needs, uses the appropriate tools, gathers the results, and produces an answer based strictly on that evidence.
For example, investigating a checkout problem might involve:
- Funnel → Identify the exact step where conversion changed
- API monitoring → Find elevated 4xx/5xx errors on
/checkout - Feature flags → Inspect recent rollout state and variations
- Users & Cohorts → Identify affected sessions and device segments
- Session replay → Surface specific recordings worth watching
You ask one question. TraceIQ handles the cross-surface investigation.
Every answer should show its work
AI answers become dangerous when confident language hides weak evidence.
We designed Ask TraceIQ around a strict, three-tier evidence classification:
- Observed — Directly supported by verified TraceIQ telemetry.
- Correlated — Two events occurred together, but the data does not establish causation.
- Hypothesis — A possible explanation that requires further investigation.
Ask TraceIQ attaches concrete evidence cards to its findings and links that evidence directly back into the relevant areas of TraceIQ.
So instead of a vague hallucination:
"Your new feature caused checkout conversion to fall."
You receive grounded, verifiable claims:
Observed: Users exposed to
checkout-v2had a lower observed checkout conversion during this period.
Correlated: API latency and 500 errors on/api/paywere elevated among affected sessions.
Hypothesis: Payment gateway timeouts may be impacting the final step of the new checkout rollout.
With direct links to inspect the flag, funnel, error issue, user timeline, or session replays yourself.
The goal isn't to replace engineering judgment. It's to make getting to the verified evidence significantly faster.
Ask from anywhere in TraceIQ
Ask TraceIQ is available globally with Cmd + K, but it also understands your active context.
- Open it while looking at a Funnel, and that funnel automatically becomes part of the investigation context.
- Open it while viewing an Error Issue, Feature Flag, API Endpoint, or Session Replay, and TraceIQ starts there instead of making you explain everything from scratch.
That makes follow-up questions instantaneous:
What changed here?
Who is affected?
Is this happening on mobile too?
What should I inspect next?
Built with strict privacy boundaries
Product telemetry can contain sensitive information, so we designed the AI layer with defensive boundaries from the beginning:
- Strict tenant isolation: Operates solely within your authenticated organization and project.
- Read-only tools: Cannot mutate data, toggle flags, or alter settings.
- Sanitized payload pipeline: Automatically strips credentials, authorization tokens, cookies, IP addresses, and sensitive parameters before sending context to the model provider.
- Untrusted data guards: Treats all event properties, error messages, and raw telemetry as untrusted data to prevent prompt injection.
- Resource bounds: Enforces deterministic limits on tool calls, token budgets, and investigation duration.
A question asked in one project cannot inspect, leak, or bleed into another.
One investigation, not one model call
A useful product investigation often requires several steps.
Ask TraceIQ may need to inspect a funnel, compare an event trend, check error occurrences, and retrieve relevant sessions before synthesizing an answer.
We treat all of that as one investigation.
Internal tool calls, reasoning rounds, and intermediate steps do not individually consume your allowance.
- Free workspaces include 3 Ask TraceIQ investigations each month.
- Pro workspaces include 100 Ask TraceIQ investigations each month.
Furthermore, we only count successfully completed investigations. Provider failures, timeouts, insufficient data, and rejected requests do not consume your allowance.
This is only the beginning
Today, Ask TraceIQ lives inside the dashboard drawer and global command palette.
But we built the underlying architecture so the intelligence layer isn't coupled to a single interface.
The same investigation engine will soon power the places where engineering and product teams already communicate:
- Slack
- Telegram
- AI-Enriched Alerts that don't just alert “error rate exceeded 5%”, but automatically investigate what changed and attach evidence before notifying your team.
Imagine TraceIQ telling your on-call channel:
Checkout failures increased sharply in the last 30 minutes. Most affected sessions are on mobile, and 82% of those users were exposed to
checkout-v2. 14 affected session replays and error traces are ready for review.
That is where product intelligence is heading.
For now, you can open TraceIQ, press Cmd + K, and ask your first question.
Your product already has the answers buried in its telemetry. Ask TraceIQ helps you find them.