Security and deployment

Keep private knowledge inside a boundary you define.

Start with one controlled server and one team. Keep documents, vectors, metadata, and access within your infrastructure, then add approved connections only when the workflow requires them.

Three controls

Make the knowledge boundary explicit.

Each control answers a practical deployment question: where knowledge lives, who can reach it, and which external systems may process it.

01 · Storage

Data stays inside

Run the Web portal, API, PostgreSQL metadata, and project vector storage on infrastructure you control. Keep uploads project-scoped and optionally encrypt newly uploaded files at rest with AES-256-GCM.

02 · Identity

Access stays scoped

Use administration roles, project manager or viewer memberships, and project-scoped MCP keys. REST credentials and MCP keys remain separate and are not interchangeable.

03 · Providers

Connections stay deliberate

Use built-in local embeddings or approved providers with independent endpoints and credentials. Inspect health, Prometheus metrics, usage rollups, and Harbor traces without opening the knowledge boundary.

Operating model

Define ownership before production.

ContextHarbor

Platform responsibility

  • Installation and product configuration
  • Supported knowledge and integration interfaces
  • Pilot guidance, documentation, and evaluation support
Your team

Environment responsibility

  • Infrastructure, network, and identity controls
  • Provider credentials, document ownership, and retention
  • Monitoring, approvals, and production operations

Start with a controlled boundary

One server. One team. One knowledge workflow.

Define where the knowledge lives, who can use it, and which connections are allowed before expanding the deployment.