Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mcp-server search langchain documentation for tool calling"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
This repository contains an implementation of a Model Context Protocol (MCP) server. This project demonstrates how to build and run a functional MCP server that can integrate with LLM clients like Claude Desktop.
This MCP server provides a suite of tools focused on information retrieval and vector database management, primarily leveraging LangChain and ChromaDB:
Targeted Documentation Search:
Quickly search the official documentation for popular AI/ML libraries:
Langchain
OpenAI
LlamaIndex
Retrieves relevant text snippets directly from the documentation sites.
ChromaDB Vector Database Integration:
Setup: Create and initialize ChromaDB vector stores with your own text data and optional metadata.
Persistence: Option to persist databases to disk for later use or use in-memory stores.
Querying: Perform semantic searches on your ChromaDB instances to find relevant documents based on query similarity.
Demonstration: A built-in demo tool to showcase the setup and query capabilities with sample data.
Powered by LangChain:
Utilizes LangChain for core functionalities like document handling, embedding management (using OpenAI Embeddings by default), and vector store interactions.
Related MCP server: Documentation Retrieval MCP Server (DOCRET)
Python 3.11 or higher (as specified in pyproject.toml)
uv package manager
Dependencies listed in pyproject.toml (e.g., mcp[cli], httpx, langchain)
uv Package ManagerIf you don't have uv installed, you can install it using:
curl -LsSf https://astral.sh/uv/install.sh | shRestart your terminal after installation.
Clone this repository (if you haven't already) and navigate into the project directory:
# cd /path/to/your/mcp-serverCreate a virtual environment and install dependencies:
uv venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
uv pip install -r requirements.txt # Or use uv pip install -e . if setup.py or pyproject.toml is configured for editable install
# Based on your pyproject.toml, you might also directly use:
# uv add beautifulsoup4 httpx "mcp[cli]" langchain langchain-community langchain-core chromadb
# Or more simply if pyproject.toml is complete:
# uv sync(Note: Ensure your pyproject.toml is complete or you have a requirements.txt for uv pip install -r requirements.txt. uv sync is often preferred if pyproject.toml defines all dependencies.)
To start the MCP server, run:
uv run main.pyThe server will start and be ready to accept connections.
To connect this MCP server to Claude Desktop:
Ensure Claude Desktop is installed.
Edit the Claude Desktop configuration file located at ~/Library/Application Support/Claude/claude_desktop_config.json (on macOS).
Add or update the mcpServers section:
{
"mcpServers": {
"mcp-server": { // You can choose any name
"command": "/full/path/to/your/.venv/bin/uv", // Use absolute path to uv in your venv
"args": [
"run",
"main.py"
],
"dir": "/full/path/to/your/mcp-server" // Absolute path to this project directory
}
}
}Important: Replace /full/path/to/your/... with the correct absolute paths on your system. Using the uv from your project's virtual environment is recommended.
Restart Claude Desktop.
This project is largely based on the mcp-server-example generously provided by Alejandro AO. We have adapted and utilized significant portions of his original work to build this server. We extend our sincere gratitude to Alejandro for his excellent example and for making his code available to the community.
You can find Alejandro AO's original repository here: https://github.com/alejandro-ao/mcp-server-example.
This project is licensed under the MIT License. See the LICENSE file for more details (if one exists).
Demonstrate the ChromaDB vector database with LangChain integration.
Args:
sample_texts: Optional list of sample texts to use for demonstration.
If not provided, default sample texts will be used.
Returns:
Results of setting up and querying the ChromaDB vector database.
Inputs: sample_texts.
Search the latest docs for a given query and library. Supports langchain, openai, and llama-index.
Args: query: The query to search for (e.g. "Chroma DB") library: The library to search in (e.g. "langchain")
Returns: Text from the docs
Inputs: query (required), library (required).
Query a ChromaDB vector database using LangChain integration.
Args:
query: The query string to search for in the vector database.
persist_directory: Directory path where the vector database is persisted.
top_k: Number of top results to return (default: 3).
Returns:
List of retrieved documents with their content and metadata.
Inputs: query (required), persist_directory (required), top_k.
Set up a ChromaDB vector database with LangChain integration.
Args:
texts: List of text strings to be embedded and stored in the vector database.
metadatas: Optional list of metadata dictionaries corresponding to each text.
persist_directory: Optional directory path to persist the vector database.
If None, an in-memory database will be created.
Returns:
Dict containing retriever info and success status.
Inputs: texts (required), metadatas, persist_directory.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Scored across 4 tools
Each tool has a clearly distinct purpose: get_docs searches documentation, setup_chroma_db initializes a vector database, query_chroma_db retrieves results, and chroma_db_demo runs a combined demonstration. No two tools overlap in a way that would cause misselection.
Three tools follow a clear verb_noun pattern (get_docs, setup_chroma_db, query_chroma_db), while chroma_db_demo deviates to a noun_noun style. However, all names use snake_case and are predictable and readable.
With four tools focused on docs search and ChromaDB operations, the count is well-scoped and each tool earns its place. It is neither too thin nor overloaded for the apparent purpose.
The core workflows are covered: search docs, set up a vector DB, and query it. Minor gaps exist—no update/delete for vectors and docs search limited to three libraries—but agents can work around these limitations.
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
An MCP server that integrates with Discord to provide AI-powered features.
An MCP server that gives your AI access to the source code and docs of all public github repos
MCP server for langchain documentation, generated by doc2mcp.
