Skip to content

Security: botAGI/morpheus-ai

Security

SECURITY.md

Security Policy

Supported Versions

Morpheus is pre-1.0 software. Security fixes target the current main branch and the latest tagged release.

Version Supported
0.1.x Yes
Older snapshots No

Reporting a Vulnerability

Report suspected vulnerabilities privately before opening a public issue. Until a dedicated security advisory channel is configured, email team@morpheus.ai with:

  • affected version or commit,
  • operating system and Python version,
  • reproduction steps,
  • impact and whether secrets, filesystem data, MCP/A2A traffic, or receipts are involved.

We will acknowledge actionable reports, avoid public disclosure until a fix is available, and credit reporters when requested.

Local-first Security Model

Morpheus is designed to run next to the project it compiles. Treat the selected project root, .morpheus/, WAKE.md, receipts, integration caches, and model smoke prompts as local project data.

Recommended defaults:

  • bind API and UI to 127.0.0.1 on untrusted networks,
  • use 0.0.0.0 only for trusted LAN testing,
  • put authentication, a trusted tunnel, or a reverse proxy in front of remote access,
  • do not commit .morpheus/, generated receipts, local integration caches, or tokens,
  • keep integration token files outside the project checkout,
  • run morpheus verify --all after meaningful state changes.

MCP/A2A Exposure

The MCP and A2A endpoints are intended for local agent interop and LAN tests. They should be treated as automation surfaces: any remote exposure needs network controls, request logging, and an explicit trust boundary.

There aren't any published security advisories