When self-hosting makes sense
Data residency requirements (regulated industry, sovereignty). Air-gap or restricted-network environments. Cost predictability at scale (avoid SaaS multipliers). Audit / change-control requirements that need full system access. For most orgs at <100 engineers, SaaS is the right choice; the self-host path becomes rational at higher scale or specific regulatory needs.
What the Helm chart includes
The Fastify API gateway. The Next.js web app. The Postgres schema (you bring the Postgres instance — Azure / RDS / on-prem). Drizzle migrations. The MCP server is independent (runs on engineers' laptops; talks to your self-hosted endpoint). Optional: pgvector for semantic search; Redis for rate limiting.
What you operate
Your Postgres (with pgvector if you want semantic search). Your Kubernetes cluster. Your TLS certs. Your DNS. Your monitoring (Prometheus / Datadog / whatever you use). Standard operational stack; no new tooling required.
Updates
We ship Helm chart updates monthly. You pull and apply. Updates are SemVer; major-version updates require migration runbooks (provided). Most updates are zero-downtime via rolling deploy.
Air-gap mode (v0.8)
For fully air-gapped deployments: receipts buffer locally; admin manages periodic verifiable export to the regulated record system. The MCP server runs in offline-queue mode. Useful for classified environments + GxP-validated systems.
Support model for self-hosted
Enterprise tier includes 24/7 support for self-hosted deployments. Critical-bug response SLA 4 hours; non-critical 1 business day. Quarterly health review with our SRE team. Standard enterprise support contract.