This repository is the SanHsien maintenance fork of anthropics/commerce-agents. Report vulnerabilities that affect this fork via this repository's Security tab. Issues that also exist upstream should be reported to Anthropic as well, through their own channel; this fork has no special relationship with Anthropic's security team.

commerce-agents is a reference implementation, not a production system: it ships no authentication, its MCP servers bind to loopback only, checkout never charges a real card, and every merchant write is staged until a human approves it (see docs/safety.md). A deployment built on top of it is responsible for its own authentication, authorization, and compliance.

This fork tracks upstream main and does not cut its own GitHub Releases unless a fork-only fix needs an independent version. Please reproduce issues against the current main of this repository.

Please do not open a public issue for a security problem. Instead use GitHub's private vulnerability reporting:

• Go to this repository's Security tab → Report a vulnerability.

Include: affected commit, a minimal reproduction, and the impact you observed.

• Read docs/safety.md before deploying: it lists every enforced rule, its module, and what a deployment must add on top (real authentication, business rules, compliance).
• Never commit a real ANTHROPIC_API_KEY; .env.example is a template, not a secret.
• The example MCP servers bind to loopback and have no authentication; do not expose them on a network interface without adding your own.