Releases: maxmalkin/AgentAuth
Releases · maxmalkin/AgentAuth
v1.2.0
What's new
- Crate renaming: removed
agentauth-prefix from all crates — nowcore,registry,sdk,py,schema(#12) - k6 load test suite: token-verify, token-issue, grant-request, plus composite scenarios (full-flow, mixed-traffic, 30-min soak) (#15)
- README improvements: fixed clone URL, added Bun/cargo-nextest prerequisites, CI/license/MSRV badges (#13, #14)
- Environment config: updated
.env.examplewith correct docker-compose ports and approval UI variables (#11)
v1.1.0
What's new
- Compliance test suite: 64 tests across 6 modules covering security invariants — token security, behavioral envelope enforcement, DPoP binding, capability boundaries, nonce replay prevention, and audit log integrity
- Approval UI redesign: Industrial dark theme with Tailwind CSS, skeleton loading states, staggered animations, and improved UX for capability review and two-step confirmation flows
v1.0.0
AgentAuth v1.0.0
Initial release of AgentAuth - a capability-based authentication system for AI agents.
Features
- Core Library - Protocol types, Ed25519 cryptography, capability schemas
- Registry Service - Token issuance, grant management, agent registration
- Verifier Service - High-performance token verification with Redis caching
- Approval UI - React frontend for human capability approvals with WebAuthn
- Rust SDK - Full-featured client with DPoP, rate limiting, token caching
- Python Bindings - PyO3 bindings with LangChain and AutoGen integrations
Security
- HSM key storage (AWS KMS, GCP Cloud KMS, Vault Transit)
- DPoP sender-constraint for token binding
- Nonce-based replay prevention
- Immutable audit log with hash chain integrity
- Constant-time cryptographic comparisons
Performance
- Sub-5ms p99 token verification (Redis warm)
- 10,000 verifications/second throughput target
Documentation
- Threat model covering 13 attack vectors
- Operations runbook for all alerts
- Capacity planning guide