Scratchpad MCP Server

@MikePressure

リポジトリ

基本情報

カテゴリ

ファイルとストレージ

トランスポート

stdio

作者

MikePressure

投稿者

MikePressure

Scratchpad MCP Server について

scratchpad-mcp is an MCP server that gives AI agents persistent, token-efficient storage. It solves a specific waste problem: agents constantly re-read files they've already seen, re-summarize documents they've already processed, and re-load context they've already understood. Ev

#mcp-agent-memory-storage-claude-anthropic

Scratchpad MCP Server の設定

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": {
    "scratchpad": {
      "command": "node",
      "args": [
        "dist/index.js"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "APIKEYHERE"
      }
    }
  }
}

Scratchpad MCP Server のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

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 Scratchpad Mcp?

Scratchpad Mcp is an MCP server that provides persistent, token-efficient storage for AI agents. It allows agents to store versioned files, append logs, and retrieve diffs or summaries instead of re-reading entire files, saving tokens and context space.

How to use Scratchpad Mcp?

Install Node.js 20+, clone the repository, run npm install and npm run build. Configure it as an MCP server in Claude Desktop by adding the server definition with the ANTHROPIC_API_KEY environment variable (required only for summarize_file). Invoke the eight tools (write_file, read_file, append_log, read_log, list_files, delete_file, summarize_file, get_usage_stats) with an agent_id parameter.

Key features of Scratchpad Mcp

Versioned writes with automatic 10‑version retention per file.
Line‑level diffs returned on read, not full content.
Append‑only logs with cursor‑based pagination (100 entries per page).
On‑demand file summarization via Claude Haiku, cached per version.
Per‑agent namespacing prevents cross‑agent data leaks.
SQLite backend with size caps and per‑agent quotas.

Use cases of Scratchpad Mcp

An agent tracks changes to a working document across turns.
An agent records its own action history and replays it efficiently.
An agent summarizes a long file once and retrieves the cached summary later.
Multiple agents share a server instance without reading each other's data.

よくある質問

What is the security model for Scratchpad Mcp?

`agent_id` is a plaintext tool parameter with no authentication; the server trusts the caller. It is safe for one‑user‑per‑server deployments (e.g., Claude Desktop), but **not** for multi‑tenant standby mode where untrusted callers share one SQLite file. Defense in depth includes SQL parameterization, path validation, size caps, and per‑agent quotas.

What are the runtime and dependency requirements?

Node.js 20+ and a SQLite backend (bundled). An Anthropic API key is required only for `summarize_file`; the other seven tools work without one. Optionally set `SCRATCHPAD_DB_PATH` to override the default SQLite file location.

How does storage work in Scratchpad Mcp?

All data lives in a single SQLite file with tables for files, versions, log entries, summaries, and agent usage. Each `write_file` stores the full content of the new version (not deltas); diffs are computed on read by comparing two versions. Older versions are pruned when the 10‑version cap is exceeded.

What are the operational limits?

1 MB per file write, 64 KB per log entry, 10 retained versions per file, 100 log entries per `read_log` page. Per‑agent quotas: 1000 files, 100k log entries, 100 MB total storage.

Who pays for `summarize_file`?

The caller always pays. In local installs, the user provides their own `ANTHROPIC_API_KEY`. In hosted deployments (e.g., Apify), each caller supplies their own API key per run. The other seven tools require no API key.

「ファイルとストレージ」の他のコンテンツ

Claude Document MCP Server

alejandroBallesterosC

A lightweight MCP server for processing, editing, and interacting with PDF, Word, Excel, and CSV documents.

mcp-rquest

xxxbrian

A MCP server providing realistic browser-like HTTP request capabilities with accurate TLS/JA3/JA4 fingerprints for bypassing anti-bot measures. It also supports converting PDF and HTML documents to Markdown for easier processing by LLMs.