$8M · AI + Sazabi

What they're building

Sazabi is observability rebuilt for teams that ship AI-generated code continuously. The thesis: Cursor, Claude Code, and Codex transformed how software gets written, but the second half of engineering — monitoring, debugging, incident response — is stuck in the pre-AI era of hand-configured dashboards, brittle instrumentation, and noisy alerts. Instead, Sazabi deploys AI agents that learn a team's logs, infrastructure, and codebase, then proactively detect, investigate, and resolve production issues — including opening pull requests with fixes, or launching a Cursor cloud agent to ship the change. The architecture is deliberately logs-first: founder Sherwood Callaway's view is that 'logs are just events, metrics are aggregated events and traces are basically correlated events,' so Sazabi accepts only logs (stdout/stderr, OpenTelemetry-compatible) and derives metrics and traces on the back end, with an AI-compressed storage layer that summarizes an hour of logs into a queryable package. Chat is the primary interface (Slack is the main entry point), plus a CLI that gives your coding agent visibility into observability data. Closed alpha: 50 teams onboarded in two weeks, 8,000 background investigations, 2,000 issues detected, 200 PRs opened. Integrations with 35+ hosting services including Vercel, AWS, GCP, Cloudflare, Neon, and Supabase.

Why this matters

The angel list is the signal: 60+ operators from Vercel, Cursor, OpenAI, Anthropic, GitHub, and Replit are exactly the people who see where AI-era developer tooling is going, and they put money in. J2 Ventures GP Christine Keung supplied the category-defining frame: 'If Datadog defined observability during the cloud-native era, Sazabi is defining it for the AI-native one.' Hunter Walk of Homebrew, who backed code-review platform Graphite early, said Sazabi 'reminds me of Graphite in the early days.' Callaway's own positioning: 'Sazabi is what Datadog would look like if it were built in 2026 instead of 2010.' The company deliberately targets early- and growth-stage startups rather than enterprises locked into legacy stacks, and differentiates from AI-SRE startups like Traversal and Resolve by owning its own storage layer instead of sitting on top of Datadog. The counterweight to know: Datadog and Grafana are incumbents actively adding AI features, so Sazabi is betting a rebuilt-from-first-principles product beats bolted-on AI.