Doc-lib

You run a self-contained MCP server that ingests documents, chunks content for semantic search, and manages notes. This guide helps you deploy and use the doc-lib-mcp server, configure local execution, and perform common actions like adding notes, ingesting files, and searching content.

How to use

To interact with the doc-lib MCP server, you run it locally and connect using its supported tools. You can add notes, ingest various document types, perform semantic searches over chunks, and retrieve relevant content for AI-assisted tasks. Use the provided commands to start the local server, then access the MCP client or integration you prefer to execute the available tools.

How to install

Prerequisites you need before starting: Python installed on your system, and a compatible runtime for the MCP executor (the project uses a standard MCP setup with the UV tooling). Ensure you have access to the file paths referenced in the commands below and that you can run shell commands from your terminal.

1. Clone the project directory that contains the MCP server code and its assets. 2) Install dependencies for the MCP server using the designated tool. 3) Start the MCP server using the runtime command shown in the configuration entries below.

Configuration

Two local runtime configurations are provided for running the doc-lib MCP server. Choose the one that matches your environment: development with a directory path or a published setup that runs directly from the MCP runner.

{
  "mcpServers": {
    "doclib_dev": {
      "command": "uv",
      "args": [
        "--directory",
        "/home/administrator/python-share/doc-lib-mcp",
        "run",
        "doc-lib-mcp"
      ]
    },
    "doclib_publ": {
      "command": "uvx",
      "args": [
        "doc-lib-mcp"
      ]
    }
  }
}

Notes on environment and tooling

The server relies on environment configuration for model access, vector embeddings, and data persistence as described in the environment variables section. You can customize runtimes and paths to fit your deployment topology, but ensure the command and arguments reflect the exact startup pattern shown above for your chosen mode.

Tool usage overview

The server exposes tools to manage notes, ingest content, search and retrieve chunks, and handle metadata. You can expand your usages by combining ingestion with semantic search to answer complex questions over your stored notes and documents.

Troubleshooting and tips

If you encounter issues starting the MCP server, verify that the path to the doc-lib-mcp directory exists and that the MCP runtime (uv or uvx) is installed and available in your shell. Check that the environment settings for the underlying models and embeddings are accessible to the process. For debugging, use a local inspector or logging to track the startup sequence and tool invocations.

Available tools

add-note

Add a new note to the in-memory note store with a name and content

ingest-string

Ingest and chunk a text string with optional source and tags

ingest-markdown

Ingest and chunk a markdown file from a given path

ingest-python

Ingest and chunk a Python file from a given path

ingest-openapi

Ingest and chunk an OpenAPI JSON file from a given path

ingest-html

Ingest and chunk an HTML file from a given path

ingest-html-url

Ingest and chunk HTML content from a URL with optional dynamic rendering

smart_ingestion

Extract technical content using Gemini and chunk it preserving code blocks and narrative content; then embed for semantic search

search-chunks

Perform semantic search over ingested content with optional type and tag filters

delete-source

Delete all chunks from a specified source

delete-chunk-by-id

Delete one or more chunks by id or a list of ids

update-chunk-type

Update the type attribute for a chunk by id

ingest-batch

Ingest and chunk multiple files in batch (markdown, OpenAPI, Python) from given paths

list-sources

List all sources with optional tag or semantic query filtering and top_k limit

get-context

Retrieve relevant content chunks for AI context with filtering and top_k

update-chunk-metadata

Update the metadata for a chunk by id

tag-chunks-by-source

Add tags to all chunks from a specific source, merging with existing tags

list-notes

List all stored notes and their content