🔍 MCP Log Analyzer
The MCP Log Analyzer is an AI-powered Streamlit app designed to analyze system log files, identify errors and warnings, and recommend fixes. It uses FastMCP, LangGraph ReAct agents, and Anthropic Claude LLM to build a powerful multi-agent system.
## 📁 Project Structure
├── analyzer.py # MCP server with two tools: analyze_logs & suggest_fix
├── streamlit_ui.py # Streamlit web interface
├── streamlit_client.py # MCP client invoking tools via LangGraph + Claude
├── mcp_config_2.json # JSON config for MCP server commands
├── test_model.py # Placeholder test script
├── README.md # ← You're here
├── temp/ # Temporary files
├── Test logs/ # Sample or uploaded logs
├── Screenshots/ # UI screenshots
├── .venv/ # Python virtual environment
Create Virtual Environment
python -m venv .venv ..venv\Scripts\activate
Install Dependencies
pip install -r requirements.txt
If requirements.txt doesn't exist, here are the needed packages:
pip install streamlit langchain langgraph langchain-anthropic anyio nest_asyncio pip install mcp langchain-mcp-adapters
🧪 Run MCP Tool Server
python analyzer.py
🧠 Run the Streamlit Client App
streamlit run streamlit_ui.py
🧾 Sample mcp_config_2.json
{ "mcpServers": { "LogAnalyzer": { "command": "{Your-directory}\uv.EXE", "args": [ "run", "--with", "mcp[cli]", "mcp", "run", "{Your-directory}\analyzer.py" ] } } }
📦 Log File Format
Uploaded logs should be a list of JSON objects like:
[ { "timestamp": "2025-07-26T12:30:01Z", "level": "ERROR", "component": "DataProcessor", "message": "NullPointerException in AuthService", "stack_trace": "java.lang.NullPointerException..." }, ... ]
📌 Notes
Claude API key is required in streamlit_client.py. Replace 'Your-API-Key' with your actual key.
If using TCP transport instead of stdio (recommended on Windows), modify the server and client configs accordingly.
You can customize or add new tools in analyzer.py and expose them via @mcp.tool().
🧠 Credits
Built using:
LangChain MCP :- https://github.com/langchain-ai/langchain/tree/main/libs/langchain-mcp-adapters
Anthropic Claude :- https://www.anthropic.com/
Streamlit :- https://streamlit.io/
LangGraph Agents :- https://github.com/langchain-ai/langgraph
🛠️ Future Improvements
Add support for batch analysis or CSV uploads
Save session history
Enable tool reordering / multiple MCPs
Deploy to Hugging Face / Streamlit Cloud
Screenshots :-
📃 License
MIT License