Skip to main content

Security Model

Orient is built with a privacy-first architecture. Your data never leaves your infrastructure.

Privacy by Design​

Orient's security model starts with a simple principle: your data stays on your machine. There is no Orient cloud, no telemetry endpoint, no analytics service. The software runs entirely within your infrastructure.

No Telemetry​

Orient sends zero data to any external service. There are:

  • No usage analytics
  • No error reporting to external services
  • No feature flags fetched from remote servers
  • No update checks phoning home
  • No behavioral tracking

Self-Hosted Only​

Orient does not offer a hosted version. Every deployment runs on the user's own infrastructure:

  • Local machine (development)
  • Private server (production)
  • Cloud VM you control (VPS, EC2, etc.)

Data Isolation​

Database​

All data is stored in a local SQLite database file. This includes:

  • Conversation history
  • Agent configurations
  • Scheduled messages
  • Integration metadata

The database file is owned by the Orient process and uses standard filesystem permissions.

Credential Management​

Sensitive credentials (API keys, OAuth tokens) are managed through the Secrets Service:

  • Secrets are stored encrypted using AES-256-GCM
  • Encryption keys are derived from a user-set master password
  • Secrets are decrypted in memory only when needed
  • No plaintext credentials are written to disk or logs

To manage secrets:

# Add a secret through the dashboard
# Navigate to Settings → Secrets

# Or via the CLI
orient secrets set OPENAI_API_KEY sk-...

See the Secrets guide for details.

Message Privacy​

  • Messages are stored locally in the SQLite database
  • Messages sent to LLM providers use your own API keys
  • No message content is shared with Orient maintainers
  • Conversation data is not used for training

Access Control​

Dashboard Authentication​

The Orient dashboard requires authentication. On first setup, you create a password that protects the web interface.

Tool Permissions​

Each agent can be configured with specific tool permissions:

  • Which integrations it can access
  • What actions it can perform
  • Rate limits on tool calls

This means you can create a "read-only" agent that can query data but not modify anything.

Platform Isolation​

Agents can be configured per-platform, so your WhatsApp agent may have different permissions than your Slack agent.

Audit Logging​

Orient logs all actions taken by agents:

  • Tool calls (what tool, what arguments, what result)
  • Integration accesses
  • Configuration changes
  • Authentication events

Logs are stored locally and can be reviewed through the dashboard or directly in the log files.

Open Source​

Orient is MIT licensed. The entire codebase is available at github.com/orient/orient.

  • Every line of code is auditable
  • No obfuscated modules
  • No binary blobs
  • Community-reviewed security practices

Best Practices​

  1. Use strong secrets — set a strong master password for the secrets service
  2. Restrict network access — run Orient behind a firewall, only expose the dashboard on localhost or a VPN
  3. Review agent permissions — give each agent only the tools it needs
  4. Keep updated — pull the latest version for security patches
  5. Backup your database — the SQLite file contains all your configuration and history