New integrations and data sources
- New modules: LangSmith, Sentra (data security posture), Neon, and Bitbucket Cloud.
- New data from existing modules: 1Password vault access, Tenable Web App Scanning findings, AWS WAFv2 web ACLs, and Maven dependencies from Semgrep.
- CVE metadata: now pulls in the CISA KEV catalog and Vulnrichment SSVC, so you can see which vulnerabilities are known to be exploited.
- Anthropic module: supports workload identity federation and Claude Enterprise orgs.
Connecting findings to assets
- Wiz and Orca findings link to the assets they affect.
- CrowdStrike hosts link to their EC2 instances, and Spotlight vulnerabilities are grouped by application.
- Tenable cloud assets link to AWS, Azure, and GCP instances.
- Sentra alerts link to AWS resources.
- Orca and AWS Inspector packages map to a shared
PackageVersionnode, so the same vulnerable package can be followed across tools.
Network and exposure modeling also improved:
- AKS services and ingresses link to Azure load balancers.
- Kubernetes ingress hosts link to DNS records.
- ECS tasks with no modeled path to the internet are now marked as such.
- Trivy now records severity, source, and fix status per image, plus where each package was found.
Supply chain
- Container images on ECR, GHCR, and GitLab now carry their source repo, read from standard OCI labels, which gives reliable image-to-code provenance.
- Images labeled with a monorepo folder URL now link correctly.
- Dockerfile-based matching is stricter: it no longer overrides images whose provenance is already known, and it needs stronger evidence before declaring a match.
- GCP Artifact Registry images now link to their layers. That layer graph can be synced selectively.
Performance
- EMR now syncs only clusters that aren't terminated, and IAM skips roles that haven't been used since the last report.
- AWS Inspector advisory data is stored once per vulnerability instead of once per finding, which shrinks the graph.
- Cleanup jobs and GCP Artifact Registry enrichment do much less redundant work.
Reliability
- Syncs retry instead of failing on throttling or transient errors from Microsoft Graph, Intune, AWS Inspector, SSM, IAM, and Wiz.
- Okta requests now time out instead of hanging, and queries retry when a Neo4j cluster member restarts.
- GCP stops making calls to a repository or dataset after its first 403 instead of repeating the same error.
- Syncs handle resources that disappear mid-run, such as Tailscale devices leaving the tailnet or ECR repos deleted between calls.
- Supabase skips stopped projects.
Correctness fixes
- AWS security group rule IDs now separate ports and protocol, so different rules no longer collide.
- GCP buckets no longer share a single label node by mistake.
- Slack members who switch between user and bot stay as one node.
- GitHub users who belong to multiple orgs keep their facts from every org.
- An IAM access key that was never used now has a null last-used date instead of a misleading value.
- Cleanup no longer deletes ECR or Artifact Registry layers that another account or project still uses.
- EC2 instances link only to the EKS cluster in their own account and region.
- Stale EKS OIDC providers are now cleaned up.
- A core bug where reloading a node reset properties set by analysis jobs is fixed.
Docs and housekeeping
Setup docs were updated for Tailscale, Salesforce, Wiz, and 1Password. Logs are quieter: per-item progress moved to DEBUG, and a false error on query retries was removed.
New contributors
Welcome @linted, @johan1252, and @rishi-raj-jain.
Full Changelog: 0.144.0...0.145.0