Crux Labs
Crux Labs provides an AI evaluation platform that distills expert judgment into custom eval suites and runs them continuously across 100% of production traces. The platform monitors agent behavior in real time, alerts teams to issues within a minute, and automatically opens verified, merge-ready pull requests with fixes. It is designed for engineering teams in regulated industries like healthcare, cybersecurity, and financial services.
- Artificial Intelligence
- Developer Tools
- AI Agents
- Software Only
Funding
Raised to date
$6MRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
across 1 round
AVBM$6M · Seed
Announced Oct 2026
Disclosed cumulative: $6M
Founders
Founder details are not available yet.
Product
Problem
Teams building AI agents in high-stakes industries struggle to evaluate agent behavior at scale. Manual review of a small sample of traces misses most failures, expert judgment is subjective and hard to codify, and issues often surface as user complaints or escalations weeks after they first appear.
Solution
Crux Labs provides a platform that distills expert judgment into a custom eval suite, then runs those evals continuously across 100% of production traces. The platform captures what "good" looks like from domain experts without requiring them to write rubrics, calibrates the evals to match expert scoring, and monitors every trace in real time. When a metric drops or a new failure mode appears, Crux alerts the team in under a minute via Slack, Teams, or email. The platform then identifies the root cause and opens a verified, merge-ready pull request directly in the customer's repository, enabling continuous agent improvement without manual eval engineering.
Target Audience
Primary customers are engineering and product teams at high-growth companies and Fortune 500s building AI agents in healthcare, cybersecurity, financial services, and other regulated or high-stakes industries where agent mistakes carry significant cost.
Features
- Expert knowledge capture that distills judgment from real traces and expert feedback into live evals without requiring manual rubric writing
- Custom eval suite generation with calibration to expert scores, achieving 92% agreement and a 0.82 Cohen's kappa in one example
- Real-time monitoring of 100% of production traces with alerts in under one minute via Slack, Teams, or email
- Automated root-cause analysis that opens verified, merge-ready PRs in the customer's repo with before/after performance metrics on replayed traces
- Live eval dashboards showing pass rates, drift detection, and per-criterion scoring with expert review status
- Support for regulated industries with ISO and HIPAA compliance, plus encryption at rest and in transit