Skip to content

Security: Sma-Das/synthient-mcp

Security

SECURITY.md

Security policy

Reporting a vulnerability

Use GitHub's private security-advisory flow for this repository. Do not open a public issue containing a Synthient API key, request or response body, exploit payload, private IP/domain lookup, or infrastructure detail.

Include the affected version, impact, reproduction steps with synthetic credentials, and any suggested mitigation. Maintainers should acknowledge a report before discussing disclosure timing.

Credential handling

  • API-key mode does not use a global SYNTHIENT_API_KEY; callers provide their own x-api-key header.
  • OAuth mode uses a server-side Synthient key that is separate from audience-bound MCP access tokens. Inbound bearer tokens are never passed through to Synthient.
  • Keep keys in the MCP client's environment, OS keychain, or secret manager.
  • Do not store keys in .env, .synthient, client configuration committed to Git, logs, screenshots, or issue reports.
  • Rotate and revoke a key immediately if it may have reached a public repository, log, untrusted build service, or third party.
  • Use separate, narrowly scoped keys for development and production.

The Docker build context is deny-by-default. Changes to .dockerignore or Docker COPY statements require security review so workspace credentials cannot become available to builders or image layers.

Deployment expectations

Local API-key deployments should remain bound to loopback. Remote deployments should use TLS, OAuth mode or an authenticated gateway, network restrictions, and edge rate limiting. Configure trusted-proxy CIDRs only when caller-IP forwarding is explicitly required.

ALLOWED_HOSTS and Origin validation mitigate HTTP attacks; neither replaces network access control or authentication. The optional metrics endpoint should remain private.

Supported versions

Security fixes are applied to the current release line. Consumers should deploy by immutable image digest and upgrade to the newest patched release.

There aren't any published security advisories