home / mcp / tavili date oauth mcp server

Tavili Date OAuth MCP Server

MCP_family

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "growley-mcp_tavili_date_oaut": {
      "url": "http://localhost:8001/mcp",
      "headers": {
        "JWT_SECRET_KEY": "YOUR_JWT_SECRET",
        "TAVILY_API_KEY": "your_tavily_api_key",
        "GITHUB_CLIENT_ID": "your-github-client-id",
        "GOOGLE_CLIENT_ID": "your-google-client-id",
        "GITHUB_CLIENT_SECRET": "your-github-client-secret",
        "GOOGLE_CLIENT_SECRET": "your-google-client-secret"
      }
    }
  }
}

You run a secure HTTP MCP server that combines smart web search, local AI processing, and OAuth2 authentication to let clients perform tooling calls and retrieve results in a structured way. It exposes an HTTP MCP endpoint, supports token-based access, and includes optional local AI processing and enhanced web search through Tavily and Ollama.

How to use

To use the MCP server, you run it locally and then connect with an MCP client. You can perform tool calls such as requesting the current date and time or initiating a web search that utilizes Tavily for web results, with DuckDuckGo as a fallback and Ollama for local AI processing. Access is secured via OAuth2 tokens, and you can inspect available tools and perform actions through the exposed MCP endpoint.

How to install

Prerequisites: Python 3.12 or later. Ensure you have a working Python environment and a shell to run commands.

Install the required tooling using the project’s install command.

uv sync

Optional components you can enable during setup: - Ollama for local AI processing: install Ollama, pull a model, and start the Ollama server. - Tavily API for enhanced web search: obtain an API key and set it in your environment. - OAuth2 for authentication: configure client IDs/secrets for Google, GitHub, and Microsoft as needed.

Start the HTTP MCP server when you are ready.

uv run python src/server.py

Additional information

Ollama (optional) enables local AI processing. To enable it: 1. Install Ollama from its official site. 2. Pull a model with a command like ollama pull gpt-oss:20b. 3. Run ollama serve to start the local AI service.

TAVILY API (recommended) adds powerful web search capabilities. To configure: 1. Register for a Tavily API key. 2. Set the environment variable TAVILY_API_KEY to your key. If no API key is provided, DuckDuckGo is used as a fallback.

OAuth2 authentication (optional) adds secure access. Prepare these environment variables and credentials: - JWT_SECRET_KEY for signing tokens - Google, GitHub, and/or Microsoft OAuth2 credentials if you plan to enable those providers Set these in a file or environment before starting the server.

Endpoints exposed by the HTTP MCP server include: - /mcp for primary MCP interactions - / for server information - /health for health status - /tools for available tools - /docs for automatic API documentation

Available tools

datetime_now

Returns the current date and time in ISO-8601 UTC format.

web_search

Performs a web search using Tavily as the primary source, DuckDuckGo as a fallback, and Ollama for local processing of results.