Nucleus Mcp

Name: Nucleus Mcp
Author: eidetic-works

@eidetic-works

Nucleus Mcp について

Nucleus MCP server by Eidetic Works — persistent memory and governance for Claude, Cursor, Windsurf — file-based, no vendor lock-in.

基本情報

カテゴリ

AI とエージェント

ライセンス

MIT

ランタイム

python

トランスポート

stdio

公開者

eidetic-works

投稿者

eidetic-works

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "nucleus": {
      "command": "python3",
      "args": [
        "-m",
        "mcp_server_nucleus"
      ]
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Nucleus Mcp?

Nucleus Mcp is an MCP server that provides a portable decision log (.brain folder) for AI coding tools. It stores decisions, policies, and plans in plain JSON and markdown files, giving Claude Code, Cursor, and Codex persistent memory across sessions without vendor lock-in.

How to use Nucleus Mcp?

Install via pip install nucleus-mcp or npx -y nucleus-mcp, then run nucleus init --recipe founder to auto-configure your MCP client. Add the remote server at https://relay.nucleusos.dev/mcp for no-install use with ChatGPT, Claude, or Perplexity.

Key features of Nucleus Mcp

114 MCP tools across 13 facades
GROUND: 5-tier execution verification (diff, syntax, imports, tests, runtime)
ALIGN: One-call human corrections with verdict + delta + DPO
Persistent engram memory across sessions and tools
Governance with kill switch, HITL gates, and compliance configs
Multi-brain orchestration, agent slots, and task dispatch

Use cases of Nucleus Mcp

Maintain project context across AI coding sessions without re-explaining
Automatically verify AI-generated code correctness and reliability
Record and learn from human corrections to improve future outputs
Generate compliance-ready audit trails for regulated workflows
Manage and prioritize tasks with escalation and heartbeat monitoring

FAQ from Nucleus Mcp

What makes Nucleus Mcp different from other context tools?

It is file-based (plain JSON + markdown), uses no embeddings, and has no vendor lock-in. All AI tools that support MCP share the same .brain folder.

What are the runtime requirements?

You can run it via Python (pip install) or Node.js (npx -y nucleus-mcp). No Python environment is needed if using npx.

Where is data stored?

Data lives locally in a .brain directory, discovered automatically by environment variable, walking up from the current working directory, or falling back to $HOME/.nucleus/brain.

Does Nucleus Mcp collect telemetry?

Yes, anonymous aggregate usage statistics (command name, duration, error type, versions, OS) are collected, but no engram content, file paths, prompts, API keys, or PII. Telemetry can be disabled with nucleus config --no-telemetry.

Are there any limits in the free version?

The free MIT version includes all 13 tools, 10 resources, 3 prompts, persistent memory, governance, and audit trails. Nucleus Pro ($19/month) adds signed audit reports, full compliance exports, and priority issue support.

「AI とエージェント」の他のコンテンツ

Solon Ai

opensolon

Java AI application development framework (supports LLM-tool,skill; RAG; MCP; Agent-ReAct,Team-Agent). Compatible with java8 ~ java25. It can also be embedded in SpringBoot, jFinal, Vert.x, Quarkus, and other frameworks.

416