Skip to content

Security: Monadix-AI/monad

SECURITY.md

Security Policy

Supported versions

monad is pre-1.0 and evolving quickly. Security fixes are applied to the main branch and the latest release only. Older releases are not patched — please upgrade before reporting.

Reporting a vulnerability

Please do not open a public issue for security vulnerabilities.

Report privately through GitHub's private vulnerability reporting ("Report a vulnerability" under the repo's Security tab). This keeps the report confidential until a fix is available and lets us coordinate a release with you.

When reporting, please include:

  • A description of the issue and its impact.
  • Steps to reproduce (a proof-of-concept is ideal).
  • Affected version / commit and platform (macOS, Linux, Windows).
  • Any suggested remediation, if you have one.

We aim to acknowledge reports within 3 business days and to provide a remediation timeline after triage. We're happy to credit you in the release notes once a fix ships — let us know how you'd like to be named.

Scope and threat model

monad is a local, single-user daemon. Understanding what is and isn't in scope will help you decide whether a finding is a vulnerability:

  • By default the daemon binds loopback only (127.0.0.1) plus a Unix-domain socket under ~/.monad/runtime/. Neither is reachable from another machine — a bound loopback port is not an exposed port.
  • The in-scope adversaries today are the user's own web browser (any page can reach 127.0.0.1) and the model's own tool calls. A remote network attacker is not yet in scope unless remote access is explicitly enabled.
  • Remote access (network.remoteAccess.enabled) binds 0.0.0.0 and requires a bearer token. Plain-HTTP remote access transmits that token in cleartext and is expected to sit behind TLS (reverse proxy / SSH tunnel / VPN). Reports about cleartext tokens on an unprotected http://0.0.0.0 are known and documented, not vulnerabilities.

This is an evolving posture, not a hardened one. The corresponding hardening work is tracked in docs/internal/development/security-guidelines.md. Findings that demonstrate a bypass of the documented model — or that escalate beyond it — are very much in scope and welcome.

There aren't any published security advisories