SuiAgentic MCP Server

How to use

You connect to SuiAgentic from an MCP client to supply your documents and request contextual content. Start by ensuring the MCP server is running locally or at a reachable address, then configure your MCP client to point to the server’s MCP endpoint. Use the client to embed new documents, run semantic searches, or fetch relevant content chunks to support your AI agents during tasks.

Key usage patterns include embedding documents from URLs or local files, performing semantic searches via natural language queries, and retrieving relevant content for use in your AI workflows. The server exposes an HTTP API and a web interface to make these actions straightforward. If you need to protect document access, you can rely on authentication features supported by the system.

How to install

Prerequisites: you need Python 3.8 or newer and Docker (optional for the vector database). Ensure you have network access to pull dependencies and any required fonts or runtime assets.

# 1) Clone the repository
git clone https://example.org/mcp_agent.git
cd mcp_agent

# 2) Create a Python virtual environment
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# 3) Install dependencies
pip install -r requirements.txt

# 4) Create a .env file if needed
# QDRANT_URL=localhost
# QDRANT_PORT=6333
# QDRANT_COLLECTION_NAME=documents

# 5) Start Qdrant (vector database) (Docker)
# docker run -p 6333:6333 qdrant/qdrant

# 6) Run the application
uvicorn app.main:app --reload
# or
python run.py

# 7) Open in your browser
http://localhost:8000

Configuration and operation notes

SuiAgentic can be run with a MCP client to expose a single HTTP endpoint for MCP interactions. The server accepts embedded documents from various sources and makes the content searchable through a semantic retrieval API. Authentication options protect access to protected documents.

If you need to upload documents, you can leverage the provided tooling to bulk upload PDFs and other document formats into the Qdrant collection, enriching the index with metadata for easy filtering later.

Troubleshooting and tips

If you encounter memory issues with large documents, break them into smaller files before uploading. For very large collections, process documents in batches to keep resource usage predictable. Check application logs for any errors during embedding or indexing.

MCP server example configuration

You can expose SuiAgentic as an MCP server using a local HTTP endpoint that your MCP clients can reach. The following example shows how to register the server in an MCP configuration.

{
  "mcpServers": {
    "sui_agentic": {
      "url": "http://localhost:8000/mcp",
      "type": "http",
      "args": []
    }
  }
}

embed

Embeds documents into the Qdrant collection by extracting content from URLs or local files, splitting into chunks, and generating embeddings.

retrieve

Performs semantic search over embeddings to retrieve relevant content chunks or documents based on a natural language query.

upload_documents

Bulk upload and index documents (PDF, DOCX, TXT) with optional tags and collection naming to organize embeddings.