karthiks/query_blogger_mcp_server

Query Blogger MCP Server

An MCP Server to

• only query (not modify),
• only your whitelisted blogs (and no other) that are hosted on Blogger](https://www.blogger.com/) platform.

Why Query Blogger MCP Server to read a public blog posts on Blogger platform?

• Simple configuration and you are ready to go.
• This server is designed to be secure, efficient, and easy to use with LLMs.

1. How to use this library?

1.1. Obtain Blogger API Key

• Go to Google Cloud Console.
• Create a new project (or select an existing one).
• Navigate to APIs & Services > Library.
  • Search for "Blogger API" and enable it.
• Go to APIs & Services > Credentials.
  • Click "Create Credentials" and choose "API Key".
• Copy the generated API key.
• Crucially, restrict your API key! Edit the API key:
  • Under API restrictions, select Restric key and choose Blogger API from the dropdown. This ensures this key can only call Blogger API.
  • Under Application restrictions, you could optionally add HTTP referrers if your Core Agent will always call from a specific domain, or "IP addresses" if your MCP server will always run on a specific IP.
    • For most internal enterprise uses, API restrictions might be sufficient, but referrers or IP restrictions add another layer.

1.2. Setup Query Blogger MCP Server

1.2.1 Clone the Repository

```bash
git clone https://github.com/karthiks/query_blogger_mcp_server
```

1.2.2 Container-less Approach: In host machine having Python installed

• Set up Environment:

  Prerequisites: pyenv is installed and is used to install python version 3.10.16.

  cd query_blogger_mcp_server
  
  python -m venv .venv
  pyenv versions
  pyenv local 3.10.16
  pyenv version # python --version
  python -m venv .venv
  source .venv/bin/activate
  
  pip install .
  # pip install query_blogger_mcp_server
  

• Configure this MCP Server

  #Copy .env.example to .env:
  cp .env.example .env
  
  # Edit the .env file and fill in BLOGGER_API_KEY, ALLOWED_DOMAINS, and optionally adjust MCP_SERVER_NAME, MCP_SERVER_VERSION, UVICORN_HOST, UVICORN_PORT.
  
  # On Windows (in cmd or PowerShell before starting server.py)
  #$env:BLOGGER_API_KEY="YOUR_ACTUAL_BLOGGER_API_KEY"
  # On Linux/macOS
  #export BLOGGER_API_KEY="YOUR_ACTUAL_BLOGGER_API_KEY"
  

  The end-user simply modifies the .env file (or sets environment variables directly in their deployment pipeline) to configure your MCP server, providing a much smoother and more secure experience.

• Run this MCP Server

  # Run MCP Server in STDIO mode
  python -m query_blogger_mcp_server.server --stdio
  
  # Run MCP Server in HTTP mode
  python -m query_blogger_mcp_server.server
  
  # Or, the production-ready way if you have `query_blogger_mcp_server` installed:
  # uvicorn query_blogger_mcp_server.server:mcp.app --host 0.0.0.0 --port 8000 --workers 1 # Add workers for production
  

  This will start your MCP server, typically on http://0.0.0.0:8000.

1.2.2 Container-ized Approach: In host machine having Docker installed

• Set up Environment:

  cd query_blogger_mcp_server
  
  # Build Docker Image: This command reads your Dockerfile and creates a Docker image for your application. Wait for the build process to complete. You should see output indicating each step of the Dockerfile being executed.
  docker build -t query-blogger-mcp-server .
  
  

• Run the Docker Container to fire-up MCP Server

  docker run -d \
  --name blogger_mcp_instance \
  -p 8000:8000 \
  -e BLOGGER_API_KEY="your_actual_google_blogger_api_key_here" \
  -e ALLOWED_DOMAINS="kartzontech.blogspot.com,blog.codonomics.com" \
  query-blogger-mcp-server
  

  • docker run: The command to create and run a container.
  • -d: Runs the container in "detached" mode (in the background), so your terminal remains free.
  • --name blogger_mcp_instance: Assigns a unique name blogger_mcp_instance to your running container. This makes it easier to refer to later (e.g., for stopping or viewing logs).
  • -p 8000:8000: This is the port mapping. It maps port 8000 on your host machine (your computer) to port 8000 inside the Docker container.
  • -e BLOGGER_API_KEY="...": This sets an environment variable named BLOGGER_API_KEY inside the container.
  • query-blogger-mcp-server: This is the name of the Docker image you built in the previous step.

  Your MCP server should now be running inside a Docker container, accessible via http://localhost:8000 from your host machine!

• Verify if Docker Container is up and running

  docker ps
  # You should see a list of running containers, and blogger_mcp_instance should be among them, with 0.0.0.0:8000->8000/tcp in the PORTS column.
  

• View Container Logs (For debugging/monitoring)

  docker logs blogger_mcp_instance
  

• Stop and Remove the Container (When finished)

  docker stop blogger_mcp_instance
  docker rm blogger_mcp_instance
  

  When you're done, it's good practice to stop and remove the container to free up resources.

1.3. Configure Your Core Agent (LLM System)

Your Core Agent (e.g., Ollama-deployed Qwen2.5 application) would then be configured to:

• Point to your MCP Server's URL (http://your-mcp-server-ip:8000).
• Be provided with the tool definitions that FastMCP automatically generates for get_blog_info_by_url and get_latest_posts_by_blog_url.

The LLM will then learn to call these tools when a user asks about blog content.

1.3.1 Integrating locally hosted Query Blogger MCP Server to Claude Desktop

Claude Desktop as yet doesn't seem to support HTTP Transport integration with MCP Server and so I show integration configuration to local MCP server run in STDIO mode

Your claude desktop config file claude_desktop_config.json will have below snippet

{
  "mcpServers": {
    "query_blogger": {
      "command": "wsl",
      "args": [
        "bash", "-c",
        "cd /absolute/path/to/your/query_blogger_mcp_server/in/wsl && source .venv/bin/activate && python -m query_blogger_mcp_server.server --stdio"
      ]
    }
  }
}

After changing this config file, restart your claude desktop to check if the query_blogger tool is integrated successfully as shown in pics below:

If this tool is enabled, you can check the tool in action as shown in screenshot below:

Links in screenshots of conversation leveraging Query Blogger MCP below:

• QnA 1
• QnA 2

2. Testing with Your LLM:

When you send prompts to your LLM that's connected to this MCP server:

• Allowed Domain: "Can you tell me about the blog at https://blog.codonomics.com?"

  • Expected: LLM calls get_blog_info_by_url to your MCP server. MCP server calls Blogger API, returns info. LLM generates a summary.

• Allowed Domain, Posts: "What are the latest posts on https://blog.codonomics.com?"

  • Expected: LLM calls get_latest_posts_by_blog_url to your MCP server. MCP server calls Blogger API, returns posts. LLM summarizes.

• Disallowed Domain: "What's the news on https://disallowed-website.com?"

  • Expected: LLM might still call get_blog_info_by_url but your MCP server's _is_allowed_domain check will trigger. It will return an error message: "Access denied: This tool can only query blogs from pre-approved domains". The LLM should then relay this message to the user.

• Modification Attempt: If you accidentally exposed a create_post tool, and the LLM tried to use it, the Blogger API itself would likely reject the API Key (as it's only for read-only access), and your MCP server would return an error. But by only defining read-only tools, you preempt this.

Demo Client LLM Agent Leveraging this QueryBloggerMCPServer

# Terminal 1
# First Start MCP Server
cd query_blogger_mcp_server
source .venv/bin/activate
python -m query_blogger_mcp_server.server

# Terminal 2
# Pre-requisite: You have qwenqwen2.5:0.5b model installed via Ollama
ollama run qwen2.5:0.5b

# Terminal 3
# Start LLM Agent that calls the MCP server
python -m query_blogger_mcp_server.server
cd query_blogger_mcp_server/demo/
source .venv/bin/activate
python ollama_mcp_agent.py
# Prompt:
# YOU: Tell me about our blog?
# AI: **Summary:** The provided blog is called **Codonomics** and it was published on April 26, 2010. It's an online platform focused on democratizing knowledge and experience.

Development & Contribution

Thank you for considering to the development efforts of Query Blogger MCP Server!

I would encourage the community to post Issues and create Pull Requests.

Before you get started, please see our .

ToDos

[X] MVP: Allow Single Domain
[X] Allow Multiple Domains
[X] User Oboarding with README.md
[X] User onboarding with demo client llm agent leveraging this mcp server
[X] Dev Onboarding with README-DEV.md
[X] Externalize configurations
[X] Containerize with Docker
[X] Integration with Claude Desktop
[ ] Security Fix: Upgrade to Vulnerable free Docker Image version (3.10 images have 2 critical vulnerabilityes) - Python-v3.14??
[ ] Unit Test Cases
[ ] Features
    - [X] Get Blog Information by URL (get_blog_info_by_url)
    - [X] Get Latest Posts by Blog URL (get_latest_posts_by_blog_url)
    - [] Get Comments for a Specific Blog Post (get_post_comments)
        - Prompt: Are there any comments on the post, "The Truth About AI Coding Agents"?
        - Prompts: What are the comments on the blog post titled "The Truth About AI Coding Agents"?
        - Prompt: Summarize the feedback from the comments section of the article "The Truth About AI Coding Agents"