EXboys/evotown

What it solves

Evotown provides a centralized control plane for managing AI agents across an enterprise. It solves the problem of "fragmented agent runtimes" where employees run different agent frameworks (like OpenClaw or Hermes) locally on their machines, leaving IT departments without a way to unify model access, distribute tools, manage knowledge, or monitor usage and costs.

How it works

Evotown acts as a middle platform that sits between the agent runtimes and the LLM providers. It does not replace the local runtimes but instead provides a set of services that the runtimes connect to via simple configuration:

• Model Gateway: An OpenAI-compatible gateway (backed by LiteLLM) that handles identity, quotas, audit logs, and cost attribution.
• SkillHub: A private marketplace for distributing signed tool/skill packages to employees.
• Knowledge Connectors: Integration with enterprise knowledge bases (e.g., Feishu, Yuque) to provide a unified search and citation system.
• Observability: A central console for tracking agent runs, conversations, and risk events.
• Evolution Arena: An optional visual simulation environment for benchmarking and reproducible evolution experiments.

Who it’s for

IT administrators and enterprise AI teams who need to govern the rest of the organization's AI agent usage without locking employees into a specific runtime or vendor.

Highlights

• Runtime-Neutral: Works with OpenClaw, Hermes, SkillLite, or any custom agent framework.
• Self-Hosted: Designed for private deployment to maintain data privacy.
• One-Click Deployment: Includes scripts for rapid enterprise rollout and MDM integration.
• Private Skill Distribution: Allows IT to review and distribute approved agent capabilities (Skills) centrally.
• Cost & Risk Management: Provides per-key monthly token limits and burst RPM limits to prevent cost overruns.