Release 0.1.0

AgentGate v0.1.0 — Ship AI. Not Incidents.

The release gate for AI agents.

AI agents now call tools, access internal systems, and make real decisions. But most teams still ship agent updates by changing a prompt and hoping nothing breaks.

AgentGate stops unsafe agent versions before they reach production.

---

What AgentGate Does

• Auditable release reports — every release check produces a full evidence-backed report that auditors, reviewers, and compliance teams can verify offline
• Configurable release policy — define your own thresholds, tool risk levels, and role permissions per agent; the gate enforces your rules, not ours
• Deterministic ship-or-block decisions — no LLM decides your release; the verdict comes from your metrics and your policy
• Self-improving safety loop — when a release is blocked, AgentGate turns the failure into a future release control so the same issue cannot slip through again
• Works with any agent — plug in your own agent config; AgentGate is not locked to one product or domain
• AI-powered investigation — review agents explain risk patterns and suggest follow-up actions, but they never make the final call