# Sazabi raises $8M seed to make observability self-healing > Source: > Published: 2026-06-30 00:12:09+00:00 [Sherwood Callaway (@shcallaway)](https://x.com/shcallaway?ref=runtimewire) has raised an $8 million seed round for [Sazabi](https://www.sazabi.com/?ref=runtimewire), the AI-native observability startup he started after years working inside the infrastructure stack at Brex and Crunchbase. A June 29 [Aligned News post on X](https://x.com/ml_angelopoulos/status/2071629884775100648?ref=runtimewire) described the round as financing for "self healing observability" that autonomously fixes production issues in AI-agent-generated code. [Aligned News on X](https://x.com/ml_angelopoulos/status/2071629884775100648?ref=runtimewire) The underlying financing was announced on June 25 in Sazabi's own [company blog post](https://www.sazabi.com/blog/seed-round?ref=runtimewire), which confirms the currency, round type and investor lineup: an $8 million seed co-led by [J2 Ventures](https://www.j2vp.com/companies?ref=runtimewire), [Village Global](https://www.villageglobal.com/?ref=runtimewire) and [Y Combinator](https://www.ycombinator.com/companies/sazabi?ref=runtimewire), with participation from [Orange Collective](https://www.orangecollective.vc/?ref=runtimewire) and more than 60 angels. The timing matters because Sazabi is not positioning itself as another dashboard vendor with an AI chat box on top. Callaway is making a narrower and more consequential bet: if coding agents make software teams ship faster, then the bottleneck shifts from writing code to operating it. Sazabi wants to sit in that second half of the lifecycle, watching logs, codebases and infrastructure, then detecting, investigating and, in some cases, opening fixes against production problems. ### A founder selling his own scar tissue Callaway's founder-market fit is the center of the story. In Sazabi's announcement, he writes that he has spent the last 10 years in infrastructure, DevOps and observability roles at high-growth startups including Brex and Crunchbase ([company announcement](https://www.sazabi.com/blog/seed-round?ref=runtimewire)). [Y Combinator's profile](https://www.ycombinator.com/companies/sazabi?ref=runtimewire) of Sazabi says the company was directly inspired by his experience at Brex, where he helped start the infrastructure and observability engineering teams. That background also explains the product's sharpest opinion: Callaway thinks observability has become too heavy for the way AI-native teams now build. In a [LinkedIn launch post](https://www.linkedin.com/posts/sherwoodcallaway_introducing-sazabi-the-ai-native-observability-activity-7424859333861068801-Eqxn?ref=runtimewire) earlier this year, he wrote that after years working on observability systems at Crunchbase, Brex and 11x, he was tired of the sprawl: dashboards, monitors, session recordings, error tracking, APM, RUM, instrumentation and OpenTelemetry. His conclusion was not that engineers needed another panel. It was that "the best UX for observability is chat." Sazabi's seed announcement expands that thesis into three product principles: less is more, logs are all you need, and monitoring is dead ([seed post](https://www.sazabi.com/blog/seed-round?ref=runtimewire)). The second claim is the most provocative. Sazabi says it can use log data as the primary source of truth and reconstruct the views engineers usually get from metrics and traces. That is a direct challenge to how most modern observability stacks are sold and implemented. ### The product is aimed at the post-Cursor production problem Sazabi's own site shows the company moving beyond passive monitoring. In one product example, a user asks Sazabi to tell Cursor to increase the timeout on an API lambda; Sazabi replies that it has launched a Cursor cloud agent and points to a pull request opened by a Sazabi bot ([product example](https://www.sazabi.com/?ref=runtimewire)). The example is a product demo, not independent proof of autonomous remediation in live customer systems, but it shows where Callaway wants the boundary to move: from alerting humans to dispatching agents. That makes Sazabi adjacent to two crowded markets without fitting cleanly into either. It is not simply an LLM observability product for monitoring AI model behavior. Sazabi says it is a generalized observability platform for any workload, including agents. It is also not simply an AI SRE overlay that sits on top of Datadog or Grafana; Sazabi says its differentiation is vertical integration across the interface, the agent and the storage layer. The company is explicit about the incumbents it wants buyers to compare it against. On its [Y Combinator launch page](https://www.ycombinator.com/companies/sazabi?ref=runtimewire), Sazabi names Datadog, Sentry, Grafana and Axiom as legacy observability platforms, and separates itself from LLM observability companies such as Arize, Braintrust, LangChain and Raindrop. That framing is useful, but still company-supplied. The harder question is whether Sazabi can replace existing telemetry workflows for teams that already have compliance requirements, incident-review processes and years of operational muscle memory built around established tools. ### The numbers are early, and self-reported In its announcement, Sazabi says that since launching less than a month before the June 25 post, it signed up 50 teams, including Mintlify, Daytona, Mastra and Sandstone. The company also says it ran 8,000 background investigations, detected 2,000 issues and opened 200 pull requests against customer repositories during its early usage period ([company blog](https://www.sazabi.com/blog/seed-round?ref=runtimewire)). Those figures point to demand from fast-moving engineering teams, but they are not the same as retention, revenue, resolved production incidents or proof that customers are ready to let AI agents remediate serious outages without human review. The public materials also leave several financing details unstated. Sazabi disclosed the round size, stage and lead investors, but not valuation, ownership sold, revenue, annual recurring revenue or total capital raised to date. The company said the funding will go toward hiring, product development, go-to-market expansion and deeper integrations across cloud and developer platforms. It also said a self-serve offering is planned for later this year. The investor list is unusually tuned to the company's wedge. Named angels include Harrison Chase of LangChain, Merrill Lutsky of Graphite, Hunter Walk of Homebrew, Matt Biilmann of Netlify, Paul Klein of Browserbase, Ivan Burazin of Daytona and Abhi Ayer of Mastra, with additional individuals from Vercel, OpenAI, Anthropic and Replit ([PR Newswire](https://www.prnewswire.com/news-releases/sazabi-raises-8-million-seed-round-to-build-the-ai-native-observability-platform-for-fast-moving-engineering-teams-302810085.html?ref=runtimewire)). For an observability startup, that lineup is not just social proof. It is distribution into the exact developer-tooling ecosystem where coding agents, automated pull requests and AI-heavy workflows are becoming normal. ### The unresolved trust problem Sazabi's promise is warm to engineering teams because it attacks one of the least-loved jobs in software: being on call with too many dashboards and too little context. But the same thing that makes the pitch compelling also makes it risky. Detecting an issue is one level of trust. Diagnosing it is another. Opening a pull request is another. Applying a production fix without human direction is a different class of operational authority. Callaway acknowledges that tension in the company blog post, comparing Sazabi's self-healing software vision to self-driving cars: technically difficult, useful if it works and unsettling because it asks the user to take their hands off the wheel. That is the real seed-stage bet. The $8 million does not prove Sazabi can safely automate production reliability. It buys Callaway time to show that observability can become active software, not just a reporting layer. If AI coding agents keep pushing more code into production faster than teams can inspect it, the company that earns trust at the moment something breaks will own one of the more important control points in the AI software stack.