Skip to content

Security: mmsk2007/helloagi

Security

docs/security.md

Security Model

HelloAGI is local-first by default.

Core rules

  • SRG evaluates every tool call before execution.
  • The HTTP API binds to 127.0.0.1 by default.
  • API access is protected by HELLOAGI_API_KEY, and onboarding can generate that shared auth token during setup.
  • Secrets are loaded from process environment first, then local .env.
  • helloagi.onboard.json stores non-secret onboarding state only.

Channel safety

  • Telegram and Discord are optional extensions.
  • Missing tokens or missing libraries are surfaced through helloagi extensions doctor and helloagi health.
  • Multi-user memory is scoped by principal identifiers per channel.

Runtime safety

  • Use helloagi doctor to inspect local runtime readiness.
  • Use helloagi health to inspect config, storage, providers, service state, and extension readiness.
  • Use reviewer-safe policy packs for read-only workflows: helloagi run --policy reviewer.

Native service defaults

  • helloagi service install keeps the service local and authenticated.
  • Linux uses systemd --user.
  • macOS uses launchd.
  • Windows uses a user-level Scheduled Task strategy.

Before exposing HelloAGI remotely, add a reverse proxy, TLS, and explicit network policy.

There aren't any published security advisories