Agyn

agyn is an open-source, Kubernetes-native AI agent management platform for shipping agents safely across a company. It is aimed at teams that need to deploy agents into private or enterprise environments, control what those agents can access, and keep usage governed as adoption grows.

The product focuses on the operational and security side of agent deployment. It supports deploying agents into VPCs, VPNs, and private networks so they can reach internal services behind corporate firewalls. Agents are defined in code and scoped to sandboxes, policies, tools, MCPs, skills, and prompts, which helps teams manage consistent behavior across environments.

agyn also provides least-privilege controls and a policy gate that inspects tool calls before they run. The page describes static policies, a policy agent, secrets vault behavior, prompt-injection defense, human escalation for ambiguous or high-risk actions, and audit logs for governed access. It includes per-agent tracking with cost attribution, usage alerts, and budget limits so teams can monitor spend by agent and by team.

The interface shown on the homepage includes enterprise workflows for access control, team sharing, role-based permissions, live policy outcomes, and cost and usage dashboards. The site also presents Terraform-based configuration examples and a GitHub-based setup flow, suggesting it is designed for infrastructure-minded teams that want to manage agents alongside other deployment assets.

Key features:
- Deploy agents into VPCs, VPNs, private networks, and environments behind firewalls.
- Enforce least-privilege access with static policies and a policy agent that checks tool calls.
- Hide secrets from the model and defend against prompt injection and sensitive data leaks.
- Track cost by agent, team, and workflow with budgets, alerts, and cost attribution.
- Share agents across teams with role-based access control and audit logs.
- Define agents, sandboxes, tools, MCPs, skills, and prompts in git or Terraform.
- Run as an open-source, Kubernetes-native platform with GitHub-based setup and documentation.