vnem install

product update

MCP, Skill, and one-command repo install.

vnem now ships a clean-folder curl install, a local read-only MCP server, a bundled Codex skill, and a wider registry for agentic coding decisions.

What changed

vnem now has a real local MCP server. It exposes the registry, search index, best-practice notes, prompt patterns, and trust metadata through read-only tools.

  • vnem_search
  • vnem_recommend
  • vnem_get_entry
  • vnem_compare
  • vnem_best_practices
  • vnem_sources

The MCP server does not install tools, edit code, call external services, collect secrets, or run the projects it recommends. It gives the agent better perception before action.

The install path is simpler

In a clean project folder, the public curl target now gives agents everything they need:

curl -fsSL https://vnem.pages.dev/i | tar -xz

That extracts a root AGENTS.md pointer plus the full read-only .vnem/ pack. When a coding agent starts in that folder, it should read the instructions and use vnem automatically.

For existing repos with their own AGENTS.md, use the local installer instead so it updates a managed vnem block:

git clone https://github.com/Ovvuhy/vnem.git
cd vnem
npm install
npm run install:project -- /path/to/project
npm run doctor -- /path/to/project

New knowledge added

This update expands the registry with new agentic infrastructure and coding-efficiency signals.

  • Source radar for official docs, registries, MCP sources, evals, and verification inputs.
  • Clude as an early cognitive memory layer and MCP server for agent memory.
  • Swarms as a Python multi-agent orchestration framework.
  • Covenant Research as watchlist-level governance research for autonomous agents.
  • Kaimo as an early agent-commerce and x402 signal.
  • ripgrep, ast-grep, ugrep, codebase-memory-mcp, Knip, and jscpd for faster and safer agentic coding workflows.
  • Phaser, PixiJS, Three.js, Babylon.js, Excalibur, KAPLAY, Matter.js, Rapier, and PlayCanvas for browser-game stack choices.

Why this matters

Agents are becoming maintainers, researchers, tool selectors, and sometimes buyers. That makes the step before action more important.

An agent should know which tools are current, which are promising but not fully reviewed, which need sensitive permissions, what safer alternatives exist, and when grep is enough versus AST or graph search.

Small, local, read-only perception for coding agents.

Current shape

  1. Install the .vnem/ pack into a repo so agents read it through AGENTS.md.
  2. Run the local read-only MCP server for tool-based search and recommendations.
  3. Install the bundled Codex skill for a cleaner agent workflow.