LoomStack
ARCHITECTURE

How LoomStack is built.

A five-layer architecture designed for enterprise-grade orchestration. Each layer has a distinct responsibility, and each is designed to be adopted incrementally.

Control Plane

Orchestration EnginePolicy EngineWorkflow SchedulerEvent Router

Orchestrates all workflow execution — composing workflows from your org config, routing between agents and humans, and triggering checkpoints at any step where risk crosses configured thresholds.

Context Layer

Knowledge GraphContext RetrievalMemory IndexerHistory Store

Maintains organizational memory — architecture, ownership, decisions, incidents. Provides versioned context snapshots to every agent at dispatch time. Stores learning outcomes from completed workflows.

Execution Layer

Agent AdaptersCI/CD ConnectorsHuman Review InterfaceTool Integrations

Executes work via connected agents and tools. Manages parallel tracks across the full lifecycle (plan, spec, code, test, security, docs, deploy agents run where dependencies allow), staged outputs, and conflict detection before results are applied.

Observability Layer

Trace CollectorCorrelation EngineProduction MonitorAnalytics

Provides complete traceability from signal to production outcome. Runs as a continuous side-channel during deployment — not a sequential step after it. Correlates production anomalies to workflow events.

Governance Layer

Audit LogRBAC EngineApproval ChainsCompliance Reporter

Records, enforces, and reports on governance across all workflow events. Immutable, tamper-evident audit trail. Approval chain state is technically enforced — workflows cannot proceed without required approvals.

Deployment options

Managed cloud service

Fully managed SaaS. Zero infrastructure to operate.

Self-hosted (Kubernetes)

Deploy in your own cloud. Full data residency control.

Hybrid

Control plane managed; data plane self-hosted.

Integration approach

LoomStack connects to your existing tools via webhooks and APIs. No changes to your repositories or CI/CD pipelines are required. Adoption is incremental — start with one workflow, expand over time.

GitHub / GitLab webhooks
Jira / Linear API
Slack API
CI/CD via APIs (GitHub Actions, CircleCI)
Observability via Datadog, Grafana, PagerDuty

Explore the architecture in depth