summarize_commandLABSCO SUMMARY
Built for the agent's context budget, not the human reader — it summarizes to keep a run alive.
The problem it targets is context overflow: a directory listing or a stack trace can fill the window and derail an agent. Here the agent gets a focused summary and a content ID, and only fetches the full text if it must. The hint parameter is the sharp tool — ask specifically for the security or API surface rather than a generic gist.
What it is
A server that summarizes large content — command results, files, directory listings, API responses — returning a concise version while storing the full text for reference.
What you get
- summarize_command executes a command and summarizes its output
- summarize_files condenses one or many files while keeping technical accuracy
- summarize_directory turns a complex directory structure into a clear overview
- summarize_text handles any other block of text
- A hint parameter focuses the summary — for example on the security surface or the API surface
- Content IDs let the agent pull the full detail only when it actually needs it
Requirements
A provider and its key — PROVIDER (ANTHROPIC, OPENAI, OPENAI-COMPATIBLE or GOOGLE) and API_KEY. Optional MODEL_ID, MAX_TOKENS (default 1024), SUMMARIZATION_CHAR_THRESHOLD (default 512) and PROVIDER_BASE_URL for OpenAI-compatible endpoints. Install with npm; the client launches build/index.js. Set MCP_WORKING_DIR so relative file paths resolve.
Setup effort
One command plus a key — npm i mcp-summarization-functions, then supply credentials
USE CASES
01
Reading a huge command output without blowing the context
Run the command through the summarizer and get the gist, with the full output kept behind an ID.
02
Getting the shape of a large file
Summarize it with a hint focused on the part you care about, like the API surface.
03
Understanding an unfamiliar directory
Turn its structure into a clear overview instead of listing every path.
04
Condensing an API response or error log
Summarize the text block and keep the detail retrievable by ID.
5 tools
Listed in project docs. Your client may show a different list.
summarize_filessummarize_directorysummarize_textget_full_contentBEFORE YOU INSTALL
What to know first
Tools you get
5 tools named in the project's documentation.
Requires
- node on PATH
- API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
npm i mcp-summarization-functionsMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"MUST_USE_summarization": {
"command": "node",
"args": [
"<YOUR_PATH_TO_SUMMARIZATION_FUNCTIONS>/build/index.js"
],
"env": {
"PROVIDER": "ANTHROPIC",
"API_KEY": "<YOUR_API_KEY>",
"MODEL_ID": "claude-3-5-sonnet-20241022",
"MCP_WORKING_DIR": "default_working_directory"
}
}
}
}
