MidOS Research Protocol
midos-mcp
MidOS MCPコミュニティライブラリ — MCPツールとしての知識、スキル、および研究。
一度インストールすれば、あらゆるAIクライアントに接続可能です。
インストール
pip install midos-mcp要件: MidOSナレッジベースがローカルで利用可能である必要があります。MIDOS_ROOT環境変数を設定し、MidOSのインストールディレクトリ(knowledge/とmodules/を含むディレクトリ)を指定してください。
export MIDOS_ROOT=/path/to/midosRelated MCP server: Code-MCP
クイックスタート
# Start server (stdio — default for Claude Code, Cursor, etc.)
midos-mcp serve
# Start with HTTP transport
midos-mcp serve --http
# Generate config for your AI client
midos-mcp config --generate claude
midos-mcp config --generate cursor
# Check health
midos-mcp healthコマンド
midos-mcp serve
好みのトランスポートを使用してMCPサーバーを起動します。
midos-mcp serve # stdio (default)
midos-mcp serve --http # Streamable HTTP on 127.0.0.1:8419
midos-mcp serve --sse # Legacy SSE transport
midos-mcp serve --http --host 0.0.0.0 --port 9000フラグ | トランスポート | ユースケース |
| stdio | Claude Code, Cursor, Cline (デフォルト) |
| Streamable HTTP | Webクライアント, リモートアクセス |
| Server-Sent Events | レガシークライアント |
midos-mcp config
サポートされている任意のMCPクライアント用に、そのまま貼り付け可能なJSON設定を生成します。
midos-mcp config --generate claude # Claude Desktop / Claude Code
midos-mcp config --generate cursor # Cursor IDE
midos-mcp config --generate cline # Cline (VS Code)
midos-mcp config --generate windsurf # Windsurf IDE
midos-mcp config --generate continue # Continue.dev
midos-mcp config --generate zed # Zed Editor
midos-mcp config --generate opencode # OpenCode
midos-mcp config --generate http # Generic HTTP client
# Write to file
midos-mcp config --generate claude -o mcp-config.jsonmidos-mcp health
サーバーの健全性を確認します:ナレッジベースの統計、ベクトルストアのステータス、依存関係など。
midos-mcp health # Human-readable output
midos-mcp health --json # JSON output for scriptsmidos-mcp keys
階層型アクセス制御のためのAPIキーを管理します。
midos-mcp keys generate --name "my-app" --tier dev
midos-mcp keys list
midos-mcp keys revoke midos_sk_abc階層: dev (無料), pro ($10/月), admin。
設定
環境変数
変数 | デフォルト | 説明 |
| 自動検出 | MidOSインストールへのパス (必須) |
|
| HTTPサーバーのバインドホスト |
|
| HTTPサーバーのバインドポート |
|
| デフォルトのトランスポート |
ルート検出
サーバーは以下の順序でMidOSナレッジベースを特定します:
MIDOS_ROOT環境変数 (推奨)自動検出: パッケージの場所から上位ディレクトリへ遡り、
CLAUDE.md+knowledge/を探すフォールバック: 現在の作業ディレクトリ
pip経由でインストールした場合、自動検出ではナレッジベースが見つかりません。以下のいずれかを行ってください:
# Option A: Clone the repo and point to it
git clone https://github.com/MidOSresearch/midos.git
export MIDOS_ROOT=./midos
# Option B: Point to an existing MidOS directory
export MIDOS_ROOT=/path/to/your/midosmidos-mcp health で確認してください。検出されたルートとナレッジの数が表示されます。
サポートされているクライアント
クライアント | 設定ファイル | トランスポート |
Claude Desktop |
| stdio |
Claude Code |
| stdio |
Cursor |
| stdio |
Cline |
| stdio |
Windsurf |
| stdio |
Continue.dev |
| stdio |
Zed |
| stdio |
OpenCode |
| SSE |
オプションの依存関係
# Vector store support (LanceDB + embeddings)
pip install midos-mcp[vector]
# Development tools
pip install midos-mcp[dev]内容
MidOSは2つの階層にわたって68個のMCPツールを公開しています:
階層 | ツール | アクセス |
Dev | 34ツール | 無料 — APIキー不要、全コンテンツ、月間500クエリ |
Pro | 68ツール | $10/月 — セキュリティ運用, AOTC, オーケストレーション, メーカー書き込み操作 |
機能
知識 — 全ドメインにわたる46,000以上の厳選されたチャンクを検索 (全コンテンツ、切り捨てなし)
スキル — 16以上の技術スタックにわたる119個の再利用可能なパターン
EUREKA — ROIが測定された383個の検証済み改善案
真実 — 50個の検証済みパッチと修正
SOTA — 140個の最先端技術の収束上限
ベクトル検索 — LanceDB + Gemini埋め込み (3072次元) によるセマンティック検索
レート制限
階層 | クエリ/月 | コンテンツ |
Dev | 500 | 全コンテンツ (切り捨てなし) |
Pro | 100,000 | 全コンテンツ + AOTC + opsパック |
無効または期限切れのAPIキーは、自動的にDev階層にフォールバックされます。
要件
Python 3.10+
FastMCP 2.x
MidOSナレッジベース (ローカル)
ライセンス
MIT
Available Tools
18 toolsagent_bootstrapA
[DEPRECATED -- use agent_handshake instead] Generic agent onboarding. Returns default config for unknown agents.
Inputs: no named parameters. Returns: Generic wrapper for non-object return types.
agent_handshakeA
Personalized agent onboarding. Declare your environment and get optimal config.
Call this FIRST when connecting to MidOS. Pass as much info as you know. Unknown fields can be left empty -- you'll get sensible defaults.
Args: model: Your model ID (e.g. 'claude-opus-4-6', 'gemini-2.5-pro', 'opus') context_window: Your context window in tokens (e.g. 200000). 0 = auto-detect from model. client: Your CLI/IDE (e.g. 'claude-code', 'cursor', 'windsurf', 'cline') languages: Comma-separated languages (e.g. 'python,typescript') frameworks: Comma-separated frameworks (e.g. 'fastapi,react') platform: Your OS (e.g. 'windows', 'linux', 'macos') project_goal: What you're working on (e.g. 'manga engine with SVG rendering')
Inputs: model, client, platform, languages, +3 more. Returns: Generic wrapper for non-object return types.
chunk_codeA
Parse code file into semantic chunks (functions, classes, methods) for better RAG retrieval.
Args: file_path: Path to code file to chunk
Inputs: file_path (required). Returns: Generic wrapper for non-object return types.
episodic_searchA
Search episodic memory for similar past experiences using vector similarity.
Args: query: Search query describing the experience/task limit: Maximum results (default: 5)
Inputs: query (required), limit. Returns: Generic wrapper for non-object return types.
episodic_storeA
Store a new episodic memory/reflection for future learning.
Args: task_type: Type of task: CODE, RESEARCH, DEBUG, REVIEW input_preview: Brief description of the input/context success: Whether the task was successful
Inputs: task_type (required), input_preview (required), success. Returns: Generic wrapper for non-object return types.
get_eurekaB
Get a specific EUREKA breakthrough document.
Args: name: EUREKA name (e.g., 'EUREKA_CACHE_SEMANTICA' or 'ATOM_001')
Inputs: name (required). Returns: Generic wrapper for non-object return types.
get_protocolC
Get a specific protocol document.
Args: name: Protocol name (e.g., 'PROTOCOL_NEURAL_LINK')
Inputs: name (required). Returns: Generic wrapper for non-object return types.
get_skillB
Get a specific skill/capability document by name.
Args: name: Skill name (e.g., 'RAG_SYSTEMS_2026_SOTA')
Inputs: name (required). Returns: Generic wrapper for non-object return types.
get_truthC
Get a specific truth patch document.
Args: name: Truth patch name (e.g., 'AGENT_MITIGATIONS_CONTEXT_OVERFLOW')
Inputs: name (required). Returns: Generic wrapper for non-object return types.
hive_statusB
Get current status of the Midos hive system.
Inputs: no named parameters. Returns: Generic wrapper for non-object return types.
list_skillsA
List all available skills/capabilities.
Args: filter: Optional filter for skill names stack: Optional stack filter, comma-separated (e.g. 'python,react'). Skills with matching compatibility.json are prioritized.
Inputs: stack, filter. Returns: Generic wrapper for non-object return types.
memory_statsA
Get statistics about the Midos memory system (LanceDB chunks, cache status).
Inputs: no named parameters. Returns: Generic wrapper for non-object return types.
pool_signalC
Signal an action to the multi-instance coordination pool.
Args: action: Action type: 'completed', 'blocked', 'claimed', 'signaling' topic: Topic/task name summary: Brief description of the action affects: Files/resources affected (optional)
Inputs: topic (required), action (required), summary (required), affects. Returns: Generic wrapper for non-object return types.
pool_statusB
Get multi-instance coordination pool status and recent activity.
Inputs: no named parameters. Returns: Generic wrapper for non-object return types.
project_statusA
Live MidOS system status + quick-start guide for your agent.
Call this anytime to get:
Real-time knowledge base stats (chunks, skills, EUREKA, truth patches)
Vector store health
Available MCP tools with usage examples
Research queue (pending topics)
Tips to get the most out of MidOS
This is your /status command. Use it to orient yourself and teach your agent how to leverage MidOS.
Inputs: no named parameters. Returns: Generic wrapper for non-object return types.
research_youtubeB
Queue a YouTube video for research. Midos will transcribe and extract insights.
Args: url: YouTube URL to research priority: Priority: 'high', 'normal', 'low'
Inputs: url (required), priority. Returns: Generic wrapper for non-object return types.
search_knowledgeC
Search the Midos knowledge base for relevant information.
Args: query: Search query (keywords or topic) max_results: Maximum results to return (default: 5)
Inputs: query (required), max_results. Returns: Generic wrapper for non-object return types.
semantic_searchA
Semantic search using LanceDB vectors (Gemini embeddings). More intelligent than keyword search.
Args: query: Natural language query (e.g., 'how to implement RAG pipelines') top_k: Number of results (default: 5) stack: Optional stack filter, comma-separated (e.g. 'python,fastapi'). Results mentioning these are boosted.
Inputs: query (required), stack, top_k. Returns: Generic wrapper for non-object return types.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 18 tools
Multiple tools have overlapping purposes. 'get_eureka', 'get_protocol', 'get_skill', and 'get_truth' all retrieve named documents with minimal distinction. Status tools like 'hive_status', 'memory_stats', 'pool_status', and 'project_status' also overlap, with 'project_status' subsuming many others. Additionally, a deprecated 'agent_bootstrap' adds confusion.
Tool names are consistently in snake_case, but the pattern is mixed: some start with verbs (get_, list_, chunk_), others with nouns (agent_, episodic_, hive_). This inconsistency makes it harder for an agent to predict tool names, though they remain readable.
With 18 tools, the server is on the higher end of the recommended range. While each tool has a distinct function, the count feels slightly heavy for a research protocol, especially given the overlaps. It could be streamlined.
The server covers onboarding, memory, knowledge retrieval, status monitoring, and research queuing, but lacks write operations for many document types (e.g., no create/update for skills, eurekas, or protocols). The research queue only has an enqueue tool, with no management or listing beyond status.
Maintenance
Related MCP Connectors
Curated knowledge API for AI agents - skill packs, semantic search, validated patterns.
Executable skill packages for agents: vetted competence, ready to run, not just documents.
Governance copilot for AI-assisted coding. 72 packs, 532 rules, proof bundles.
Search 1,700+ agent skills for Claude, Codex and Cursor, each safety-scanned, with proofs.
Related MCP Servers
- FlicenseBqualityDmaintenanceEnables comprehensive Stacks blockchain development with 30+ specialized tools for Clarity smart contracts, SIP compliance, security analysis, and performance optimization. Provides complete access to SIP standards, token development templates, security-first patterns with mandatory post-conditions, and full-stack dApp development guidance.367-
- FlicenseBqualityNot gradedmaintenanceTurns AI assistants into full-stack software engineers with 36 tools for cognitive reasoning, code validation, project scaffolding, and AI/IDE configuration generation across 130+ programming languages, databases, and frameworks.3513 npm-
- AlicenseAqualityFmaintenanceProvides access to a library of 18 specialized skills, project templates, and prompt patterns for Claude Code. It enables automated workflows for feature development, security audits, and system maintenance through multi-agent collaboration.6MIT
- AlicenseAqualityDmaintenanceProvides 634+ curated production-ready AI skills for any MCP-compatible client, covering areas like security, frontend, backend, infrastructure, AI/ML, and architecture. Enables AI agents to access best practices, patterns, and solutions across multiple domains.61MIT