Auto-Detect & Resolve Incidents
Overseer creates incidents from errors and log patterns, deduplicates by fingerprint, ranks by severity, and dispatches AI agents to investigate and fix — all automatically.
Incident Lifecycle
Error Captured
SDK detects and reports automatically
An error occurs in the browser or is detected from log analysis. The SDK flushes its 15-second DVR buffer and sends the incident payload with full context: error, stack trace, replay clip, breadcrumbs, and user info.
Storage Architecture
Incidents are stored in a two-tier system for fast access and long-term archival:
Hot Store — Redis
Recent incidents live in Redis (Upstash) with 7-day TTL. Indexed by reference ID, digest, fingerprint, and issue ID.
- Sub-millisecond reads
- Sorted sets for "recent errors" queries
- Auto-expires after 7 days
Cold Store — R2/S3
Full incident payloads (including replay clips) are permanently archived in R2. Lookup falls back when Redis TTL expires.
- Permanent retention
- Cost-effective object storage
- Automatic fallback from Redis
Alerting
Slack
Critical incidents post to your Slack channel with title, severity, affected users, and a link to the replay.
Digest emails for medium-priority issues. Instant emails for critical signals.
AI Companion
Adjusts alert intrusiveness based on your dismiss patterns. Learns what you care about.
IDE Integration (MCP)
Query incidents directly from your IDE using the Overseer MCP tools:
Start resolving incidents faster
AI-powered investigation and autofix. Free to start.