Security Architecture & Data Sovereignty
OpenBoard is engineered from the ground up for strict data confidentiality, zero cloud egress, and air-gapped enterprise environments.
100% Local SQLite Storage
All whiteboards, shapes, connectors, and versions reside solely in ~/.openboard/openboard.db on your local filesystem.
Zero Cloud Telemetry
Zero tracking cookies, zero analytics beacons, zero telemetry pings, and zero remote logging scripts.
Isolated MCP Stdio Sandbox
AI agents interact exclusively through local standard I/O (stdio) or localhost SSE. No third-party network requests are ever triggered.
1. Threat Model & Data Boundaries
Proprietary architectural blueprints—including microservice topologies, database schemas, and cryptographic keyflows—are prime targets for intellectual property theft. Storing these diagrams on multi-tenant SaaS platforms introduces significant vulnerability vectors.
OpenBoard eliminates this attack surface by executing completely within your user account boundary. The local server binds strictly to 127.0.0.1 by default, preventing external network interfaces from accessing the canvas or SQLite database.
2. Model Context Protocol (MCP) Security Controls
When AI coding assistants like Claude Code, Cursor, or Codex connect to OpenBoard, they communicate via standard input/output (stdio) pipes spawned by the parent IDE process.
- Input Validation: All 13 semantic tools parse and validate JSON payloads using strict runtime schemas.
- Atomic Transactions: Canvas mutations are wrapped in SQLite transactions with Write-Ahead Logging (WAL) to prevent data corruption.
- No Network Egress: The MCP server process initiates zero outbound network connections.
3. Air-Gapped Verification Command
You can verify OpenBoard's network isolation by running the start command with complete offline sandboxing: