Presentation

You can use the Presentation MCP Server to enable AI agents to search, extract, and assemble presentation content from embedded documents. It provides a remote HTTP API for content discovery and slide-related tasks, plus a local stdio-based MCP option for running a server alongside your tooling. This setup is designed for seamless agent integration, fast semantic search, and structured data retrieval to build polished presentations automatically.

How to use

You interact with the server by sending requests to its HTTP API endpoints. Use it to search for presentation content, extract key points, generate slide ideas, find supporting data, and gather citations. You can also run a local instance of the server via a standard input/output channel for tight integration with your tooling.

Typical workflows include: searching for relevant content for a given topic, generating a slide structure, populating slides with content and data, and appending properly formatted citations. Agents can combine results from content search, key point extraction, and data retrieval to assemble a complete presentation deck.

How to install

Prerequisites: you need Python 3.10 or higher and a Supabase project with the vector extension (pgvector). You also require an OpenAI API key for embedding generation and a Supabase service role key.

Step 1: Set up the environment and install dependencies

# Create and activate a Python virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Step 2: Configure and run the server

# Run the server directly from source
python src/server.py

# Or install as a package and run (if applicable in your setup)
document-retrieval-mcp

Additional configuration and notes

Configure environment variables to connect to your data sources and AI services. The server uses a database URL, an OpenAI API key, and optional Supabase credentials to access embeddings and content.

Security and access controls are provided through optional authentication and environment-based filtering. You can enable JWT-based authentication if your deployment requires it.

Available tools

search_documents

Search for documents using semantic similarity with configurable limits and filtering by user/session/project. Returns matching chunks with metadata and similarity scores.

get_document_context

Retrieve full document content or specific chunks by document ID, enabling detailed citation and review workflows.

list_user_documents

List all documents accessible to a user, with pagination and optional project filtering.

get_similar_chunks

Find chunks similar to a reference chunk to support content expansion and cross-referencing.