mcp-server-sample
mcp-server-sample — Minimaler MCP-Server zum Speichern und Durchsuchen von Notizen
Ein minimaler MCP-Server, der nur Notizen speichert und durchsucht. Er verbindet sich mit keinem externen System.
Der Speicherort ist nur eine Datei notes.json im selben Ordner.
Er enthält jeweils ein Beispiel für die drei MCP-Primitive.
Primitive | Wer entscheidet | Inhalt auf diesem Server |
Tools | Das Modell entscheidet |
|
Resources | Die KI-App ruft sie ab und übergibt sie |
|
Prompts | Der Benutzer wählt explizit aus |
|
Voraussetzungen
Node.js 24 oder höher (LTS. Überprüfen mit
node --version)git
Related MCP server: mcp-snippetbox
Einrichtung
git clone https://github.com/utakatano/mcp-server-sample.git
cd mcp-server-sample
npm installWenn npm install erfolgreich war, ist die Einrichtung abgeschlossen. Zu diesem Zeitpunkt wird der Server nicht gestartet.
Der MCP-Server wird von der KI-App gestartet, daher muss er nicht manuell über das Terminal ausgeführt werden.
Mit der KI-App verbinden
Claude Code
claude mcp add notes -- node /絶対パス/mcp-server-sample/index.jsWenn claude mcp list ✔ Connected anzeigt, ist die Verbindung hergestellt.
Claude Desktop
Einstellungen → Developer → „Edit Config“ öffnet claude_desktop_config.json. Fügen Sie Folgendes hinzu:
{
"mcpServers": {
"notes": {
"command": "node",
"args": ["/絶対パス/mcp-server-sample/index.js"]
}
}
}Nach dem Speichern Claude Desktop vollständig beenden und neu starten (nur das Schließen des Fensters reicht nicht).
macOS:
~/Library/Application Support/Claude/claude_desktop_config.jsonWindows:
%APPDATA%\Claude\claude_desktop_config.json
Funktionstest
「来週までにネットワーク構成を確認する」とメモして「ネットワーク」を含むメモを探してEs wird eine notes.json erstellt, deren Inhalt nach und nach wächst.
Speicherort ändern
Durch die Umgebungsvariable NOTES_FILE mit einem absoluten Pfad kann der Speicherort geändert werden.
{
"mcpServers": {
"notes": {
"command": "node",
"args": ["/絶対パス/mcp-server-sample/index.js"],
"env": { "NOTES_FILE": "/絶対パス/my-notes.json" }
}
}
}Wenn keine Verbindung zustande kommt
Zuerst das Log prüfen. Die Ursache steht meistens dort.
tail -20 ~/Library/Logs/Claude/mcp-server-notes.log # macOS
# Windows: %APPDATA%\Claude\logs\mcp-server-notes.logAchten Sie darauf, ob nach Server started and connected successfully ein Fehler auftritt.
Im Log angezeigt | Ursache | Lösung |
| Der Pfad in | Führen Sie |
| Die KI-App findet | Geben Sie die Ausgabe von |
| Abhängigkeiten nicht installiert | Führen Sie im Repository-Stammverzeichnis |
Log ist leer oder wird nicht aktualisiert | Die Konfiguration wird nicht geladen | Überprüfen Sie die JSON-Syntax (Kommas, Klammern) und beenden Sie Claude Desktop vollständig (macOS: ⌘Q) und starten Sie es neu |
Zur Unterscheidung, ob der Fehler serverseitig oder seitens der KI-App liegt, ist es am schnellsten, den Server lokal direkt aufzurufen.
printf '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{"_meta":{"io.modelcontextprotocol/protocolVersion":"2026-07-28"}}}\n' \
| node /絶対パス/mcp-server-sample/index.jsWenn ein JSON mit add_note und search_notes zurückgegeben wird, funktioniert der Server korrekt. In diesem Fall prüfen Sie die Einstellungen der KI-App (Pfad, JSON-Format, Neustart).
Maßnahmen gegen Lieferkettenangriffe
In der .npmrc sind Einstellungen enthalten, die eine Übernahme von npm-Paketen verhindern sollen. Sie wirken bei jedem npm install / npm ci.
Einstellung | Wirkung |
| Führt bei der Installation keine Lebenszyklus-Skripte von Abhängigkeiten (z.B. |
|
|
| Installiert nur Versionen, die älter als 7 Tage sind. npm wandelt dies in |
Die Abhängigkeiten sind in package.json vollständig fixiert (@modelcontextprotocol/server ist 2.0.0, zod ist 4.4.3),
und in package-lock.json mit integrity-Hash protokolliert.
Wenn Sie die Abhängigkeiten genau wie im Lockfile installieren möchten, verwenden Sie npm ci anstelle von npm install.
Wenn Sie Abhängigkeiten aktualisieren möchten, heben Sie sie bewusst einzeln an.
npm outdated
npm install @modelcontextprotocol/server@2.1.0 # save-exact により完全固定で書かれる
npm ls --all # 増えた依存を目で確認するLizenz
MIT
Available Tools
2 toolsadd_noteメモを追加するA
メモを1件保存する。打ち合わせの決定事項や、あとで思い出したいことを記録するときに使う。
| Name | Required | Description | Default |
|---|---|---|---|
| tags | No | 分類用のタグ(任意) | |
| text | Yes | 保存する本文 |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It states '保存する' which implies mutation, but it does not disclose potential side effects, permissions, or behavior on duplicates. For a simple create operation, this is minimally adequate.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences, front-loaded with the essential action. No wasted words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple tool with 2 parameters, no output schema, and no annotations, the description is fairly complete. It explains the primary use case and differentiates from the sibling. It could mention the return value but is not required.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100% with both parameters described. The tool's description adds no new information about the parameters; it only restates the purpose. Baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb '保存する' (save) and resource 'メモ' (note), and it is distinct from the sibling tool 'search_notes' which is for retrieval.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It provides explicit context for when to use: '打ち合わせの決定事項や、あとで思い出したいことを記録するときに使う' (use when recording meeting decisions or things you want to remember later). It does not explicitly exclude other scenarios, but the purpose is clear.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
search_notesメモを検索するB
保存済みのメモをキーワードで検索する。本文とタグの両方を対象にする。
| Name | Required | Description | Default |
|---|---|---|---|
| keyword | Yes | 検索キーワード |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must fully disclose behavioral traits. It states search targets both body text and tags, but lacks details on whether searches are case-sensitive, support partial matches, or have rate limits. It does not mention return format or pagination. The description adds modest value but is insufficient for a tool lacking annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single sentence with no wasted words. It is concise and front-loaded with the action and target, then expands on scope. Every phrase adds value.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool is simple (1 parameter, no output schema, no nested objects), the description is minimally adequate. It covers the search target and scope. However, it omits any mention of results behavior (e.g., whether it returns full notes or summaries) and does not compensate for missing annotations, but the low complexity reduces the burden.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% with one parameter 'keyword'. The description adds meaningful context by specifying that keyword searches both body and tags, which the schema description ('検索キーワード') does not convey. This enriches the semantic understanding beyond the schema alone.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Description clearly specifies the action (検索する), the target (保存済みのメモ), and the search scope (本文とタグの両方). It distinguishes from add_note, which creates notes. However, it does not explicitly contrast the two, leaving a slight gap in sibling differentiation.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for keyword search but gives no guidance on when to use this vs. add_note, nor does it mention any context or prerequisites for searching. No exclusion criteria or alternative scenarios are provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
2 tool updates
v1.0.0- First observed
add_note - First observed
search_notes
TDQS
Scored across 2 tools
The two tools have completely distinct purposes: one creates/ stores a note, the other retrieves notes via search. There is no overlap or ambiguity between writing and searching.
Both tool names follow a consistent verb_noun pattern (add_note, search_notes) using lowercase with underscores. The naming is predictable and clear.
With only 2 tools, the set feels minimal. While the tools cover basic create and read/search operations, the server's purpose (saving and retrieving notes) could reasonably include more tools (e.g., update, delete, list tags) to avoid being overly thin.
The domain is a personal note-taking system, but only add and search operations are provided. Missing critical operations like update, delete, and get all notes (without search) create significant gaps that would frustrate or block a typical agent workflow.
Maintenance
Related MCP Connectors
Google Keep-style notes app with an MCP server for AI agents to read/write notes.
- TaprootOAuthcom.taproothq
Persistent memory layer for AI tools. Save and recall notes across Claude and other MCP clients.
Search, read, and write your Apple Notes from ChatGPT/Claude via a local Mac agent + MCP relay.
MCP-native notes and memory for ChatGPT, Claude, and other AI tools.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables adding, getting, and listing notes through MCP, with state persisted to a JSON file in the home directory.MIT
- AlicenseNot gradedqualityCmaintenanceEnables managing notes through a simple MCP server, providing tools to add, retrieve, update, delete, and list notes for use with Claude Desktop.MIT
- FlicenseNot gradedqualityCmaintenanceEnables AI clients to create, list, and search notes stored in a local JSON file through the Model Context Protocol.-
- AlicenseNot gradedqualityCmaintenanceEnables Claude Desktop to add, retrieve, update, delete, and list notes persisted in a JSON file whose path is set via MCP_NOTES_FILE or --notes-file, with atomic, lock-protected saves.MIT