Skip to content

Security: Akshu1245/Rakshex-complete-codebase

docs/SECURITY.md

Rakshex Security

Overview

Rakshex processes API collections, AI agent telemetry, and governance controls. This document summarizes security posture for operators and customers.

Maturity: operational guidance — not a certification claim.

Authentication and authorization

  • Passwords hashed with Argon2id
  • Sessions are server-side; OAuth uses PKCE
  • Workspace-scoped RBAC; API keys are hashed at rest
  • Cross-tenant access is denied by authorization helpers (BOLA prevention)

AgentGuard runtime

  • SDK defaults to metadata_only (no prompt content)
  • Provider API keys never leave the customer process toward Rakshex
  • Gateway kill switches: workspace, project, agent
  • Fail-open vs fail-closed configurable; emergency bypass is audited and permissioned

Secrets

  • Structured logs redact passwords, tokens, API keys, cookies
  • Collection import scans for embedded credentials
  • MCP stdio commands are allowlisted; shell metacharacters blocked

Data rights

  • Export and deletion workflows anonymize or remove personal data
  • Zero-retention and local-only modes available in the SDK
  • Retention policies document telemetry and audit windows

Reporting vulnerabilities

See /.well-known/security.txt and email security@rakshex.in.

Compliance reports

Compliance dashboards map evidence to frameworks (OWASP, NIST AI RMF, ISO, SOC 2, GDPR, DPDP, EU AI Act). Reports are not certifications.

There aren't any published security advisories