YBK · Enterprise AI Asset Governance: One control plane for every AI asset: MCP, skill, agent, app

Most teams already run MCP servers, skills, agents and AI apps. What's missing is treating them as assets — one identity, one permission model, one cost ledger, one audit trail. ybkkasi does that: • Five asset types in one ledger with a lifecycle state machine • Cost attribution by department × agent × skill — attribution, not just observation • Dependency topology and blast-radius analysis • Fully self-hosted, MIT. Node 22 + node:sqlite, zero added dependencies

Hey everyone — I'm the solo maintainer of ybkkasi (元冰可). Context in one line: we turn every MCP server, Skill, Agent and AI app in a company into an asset you can govern, cost, and audit. Three things I built that I couldn't find elsewhere: 1. Five asset types in one ledger with a real lifecycle state machine (not five separate dashboards) 2. Cost attribution on department × agent × skill — attribution, not just observation. Knowing the total went up doesn't let you cut it. 3. Fully self-hosted + MIT, zero added dependencies (Node 22 + node:sqlite). Nothing to install behind your firewall — that matters more than it sounds. v1.3.0: 27 plugin packages / 84 permission points / 354 REST routes / 83 MCP tools / 1486 selftest assertions green. zh-CN, en-US, ja-JP. Repo is MIT if you'd rather just read the code: https://github.com/01men/ybkkasi I'll be in the comments all day — ask me anything, especially the hard ones.