This section was written completely by using the SentinelCore agent and its tools(prompt->get the details from internet->write it to a file) via gemini 2.0 flash.

What is SentinelCore Agent?

SentinelCore Agent is an MCP (Model Context Protocol) server that provides tools for file system operations, web scraping, AI-powered web search, and vector-based indexing. It is designed to be used with an LLM client that orchestrates tool calls, supporting both Azure OpenAI and Google Gemini models.

How to use SentinelCore Agent?

Configure the server by setting environment variables (including an LLM API key) and a JSON server configuration file. The server is started using mcp.run(transport="stdio"). The included client module (client.py) connects to the server, lists available tools, and runs a chat session where the LLM can invoke tools to answer user queries.

Key features of SentinelCore Agent

• File existence checking and read/write operations (text and binary).
• Current date and time retrieval.
• AI-powered web search via Brave Search agent.
• Web page scraping to markdown with optional vector indexing.
• Vector index management: list all indexes and search via an embedding model.

Use cases of SentinelCore Agent

• Automate file reading, writing, and metadata checks on a local filesystem.
• Scrape web pages and store their content as vector indexes for later retrieval.
• Power a conversational AI agent that can search the web and manipulate files based on natural language prompts.
• Build a research assistant that scrapes pages, indexes them, and answers queries using semantic search.

FAQ from SentinelCore Agent

What transports and authentication does the server support?

The server uses transport="stdio" and does not mention built-in authentication.

What are the main dependencies?

The server depends on fastmcp, crawl4ai (asynchronous web crawler), and an external embedding model for vector operations.

How does the client communicate with the LLM?

The client supports Azure OpenAI and Google Gemini models, configuring them via environment variables and managing tool call execution with retry logic.

Where are vector indexes stored?

Vector indexes are stored as local files in the current working directory.

Does the server have any known limitations?

The README does not list any specific limitations or rate limits.