The dashboard, page by page.

Overview

Single-screen workspace pulse: KPI strip (events / unique users / crash-free %), a 24h event timeline, top issues by event count, recent fired alerts, and the SDK health snapshot. The first thing your team sees on login.

Issues

Grouped, fingerprinted exception list. Filterable by status, severity, project, environment, and release. The search box does Meilisearch full-text matching across title, culprit, exception type, and message. The toggle next to it switches to NLQ mode — natural-language → ClickHouse SELECT via Claude.

Click any issue to see breadcrumbs, stack frames (symbolicated when source maps are uploaded), TillMind AI analysis, FrictionHeatmap (rage / dead-zone / ghost taps), comments, and a Resolve / Ignore / Assign control strip.

Performance

P50 / P75 / P95 cold-start, ANR rate, frame jank (slow >16.67ms / frozen >700ms), screen-level battery attribution, and per-transaction latency histograms. Color coded green <200ms, amber <1s, red ≥1s.

Releases

One row per (project, version, environment). Crash-free % per release with letter-grade scoring (A ≥99.5, B ≥99.0, C ≥98.0, D ≥95.0, F <95). Hourly background worker pre-computes these so dashboard reads stay cheap.

Releases → Compare shows an A/B diff between two releases — crashes, ANR rate, jank, p95 cold-start — with delta annotations.

Alerts

Two tabs: Fired (recent alert events with ack control) and Rules (alert rule editor). KPI strip across the top — total fired, unacknowledged, MTTA. See Alerts & integrationsfor the trigger/action matrix.

Ask Pulse

Multi-turn chat with Claude. Streams via SSE, renders annotated SQL with bar-chart sparklines next to each result block. The model is sandbox-only — SELECT-only, table allow-list, project filter injected server-side, never executes the SQL it generates.

TillMind

AI-generated hypothesis cards for spikes, regressions, and unusual patterns. Each card has a confidence bar and a CTA to open it in Ask Pulse for a deeper dialogue. Backed by buildHypotheses against ClickHouse.

Field map

Interactive D3-geo orthographic globe showing event distribution by country / region. Powered by fetchGeoDistribution from ClickHouse, topojson via the public CDN. Drag to rotate. Click a region to filter the rest of the dashboard.

Devices

Device, OS, and network rollup from device_stats_daily. Crash-free % per device family. Surfaces "your worst phone" — the Tecno Spark 10 on 2G that ships with 90% of crashes.

Security

Filtered view of security events: rooted device, jailbreak indicator, overlay attack, certificate-pinning failure, tamper detection. Click an event type to narrow to that pattern.

Threat intel

Cross-customer hashed indicator board. An indicator is surfaced once it is confirmed malicious, corroborated across multiple organizations, or seen locally. No plaintext indicator values stored — sha256 only.

SDK health

Capture vs. delivery rates per-project, per-platform: events captured, events delivered, dropped (queue overflow), failed (transport error). Snapshots taken on init + on flush.

Onboarding

Five-step state machine. Auto-launches for new orgs:

1. Create your first project
2. Install an SDK
3. Wait for first event (polls every 3s)
4. Configure your first alert
5. Invite your team

Settings

Tabs:

• General — org name, slug, default region.
• Projects — DSN, rotate, alert thresholds, region.
• Symbols — JS source maps / Android ProGuard / iOS dSYM uploads, sortable by release.
• Integrations — Slack, GitHub, Linear, Jira, PagerDuty, MS Teams.
• SSO — OIDC issuer, client ID, client secret, allowed domains, JIT toggle, default role.
• API keys — list, create, revoke (raw key shown once).
• Audit log — paginated, filterable; admin / owner only.

Profile

Personal: change password, enable / disable TOTP, manage personal API keys.

Keyboard shortcuts