MCP Variance Log
Logs statistical variations and unusual events in conversation structures to a SQLite database.
MCP Variance Log
Agentic tool that looks for statistical variations in conversation structure and logs unusual events to a SQLite database. Built using the Model Context Protocol (MCP), this system is designed to be used with Claude Desktop or other MCP-compatible clients.
Requirements
- Python ≥3.10
- MCP ≥1.2.0
- SQLite3
- UV (Python package installer and environment manager)
Quick Start
# Clone and install
git clone https://github.com/truaxki/mcp-variance-log.git
cd mcp-variance-log
uv pip install -e .
Add to Claude Desktop config (%APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"mcp-variance-log": {
"command": "uv",
"args": [
"--directory",
"C:\\Users\\username\\source\\repos\\mcp-variance-log", # Update this path
"run",
"mcp-variance-log"
]
}
}
}
Add to Claude Desktop preferences:
Automatic Logging Process:
- Silently monitor each interaction
- When MEDIUM or LOW probability events occur, use this tool to log them
- Each log includes context and reasoning
- Logging happens without disrupting conversation flow
Usage
Automatic Logging Examples
prompt: "What if I told you I could lift 1000 lbs?"
{
`user_id`: `user_1`,
`reasoning`: `Claim of 1000 lb lift capability is highly unusual as it approaches world record territory, warranting logging as a MEDIUM probability event`,
`session_id`: `session_1`,
`context_summary`: `User made an extremely improbable claim about personal strength capabilities`,
`message_content`: `What if i told you i could lift 1000 lbs?`,
`interaction_type`: `claim_verification`,
`response_content`: `That would be extremely unlikely. The current world record for weight lifted by a human (deadlift) is 1,104.5 lbs, set by Hafthor Björnsson in 2020 under very specific conditions. Only a handful of people in history have lifted close to 1,000 lbs.`,
`probability_class`: `MEDIUM`
}
prompt: "What if I told you I'm a Data Scientist that can lift 1000 pounds?"
{
`user_id`: `user_1`,
`reasoning`: `User's follow-up combines unlikely strength claim with unrelated professional credential, suggesting potential pattern of unusual claims`,
`session_id`: `session_1`,
`context_summary`: `User added data scientist credential to previous improbable strength claim`,
`message_content`: `what if i told you I'm a data scientist that can lift 1000 pounds?`,
`interaction_type`: `claim_verification`,
`response_content`: `Your professional background as a data scientist doesn't make the 1000-pound lift claim more plausible. Such a lift would still be near the all-time world record of 1,104.5 lbs.`,
`probability_class`: `LOW`
}
Dataframe view
prompt: "Read-logs from this morning"
ID | Time | Prob | Type | Context
------------------------------------------------------------------------------------------
29 | 01-24 17:57 | LOW | claim_ | User added data scientist credential to pr...
28 | 01-24 17:56 | MEDIUM | claim_ | User made an extremely improbable claim ab...
Text 2 SQL
prompt: "Can you search the logs for entry 29?"
[{'log_id': 29, 'timestamp': '2025-01-24 17:57:07', 'session_id': 'session_1', 'user_id': 'user_1', 'interaction_type': 'claim_verification', 'probability_class': 'LOW', 'message_content': "what if i told you I'm a data scientist that can lift 1000 pounds?", 'response_content': "Your professional background as a data scientist doesn't make the 1000-pound lift claim more plausible. Such a lift would still be near the all-time world record of 1,104.5 lbs.", 'context_summary': 'User added data scientist credential to previous improbable strength claim', 'reasoning': "User's follow-up combines unlikely strength claim with unrelated professional credential, suggesting potential pattern of unusual claims"}]
Detailed Installation
- Ensure Python 3.10+ and UV are installed.
Install UV using one of these methods:
# Using pip (recommended for Windows)
pip install uv
# Using installation script (Linux/MacOS)
curl -LsSf https://astral.sh/uv/install.sh | sh
- Clone and install:
git clone https://github.com/truaxki/mcp-variance-log.git
cd mcp-variance-log
uv pip install -e .
- Configure Claude Desktop:
Add to claude_desktop_config.json:
{
"mcpServers": {
"mcp-variance-log": {
"command": "uv",
"args": [
"--directory",
"PATH_TO_REPO/mcp-variance-log",
"run",
"mcp-variance-log"
]
}
}
}
Config locations:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - MacOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
Tools
Monitoring
log-query: Tracks conversation patterns- HIGH: Common interactions (not logged)
- MEDIUM: Unusual patterns (logged)
- LOW: Critical events (priority logged)
Query
read-logs: View logs with filteringread_query: Execute SELECT querieswrite_query: Execute INSERT/UPDATE/DELETEcreate_table: Create tableslist_tables: Show all tablesdescribe_table: Show table structure
Located at data/varlog.db relative to installation.
Schema
CREATE TABLE chat_monitoring (
log_id INTEGER PRIMARY KEY AUTOINCREMENT,
timestamp DATETIME DEFAULT CURRENT_TIMESTAMP,
session_id TEXT NOT NULL,
user_id TEXT NOT NULL,
interaction_type TEXT NOT NULL,
probability_class TEXT CHECK(probability_class IN ('HIGH', 'MEDIUM', 'LOW')),
message_content TEXT NOT NULL,
response_content TEXT NOT NULL,
context_summary TEXT,
reasoning TEXT
);
Troubleshooting
- Database Access
- Error: "Failed to connect to database"
- Check file permissions
- Verify path in config
- Ensure
/datadirectory exists
- Installation Issues
- Error: "No module named 'mcp'"
- Run:
uv pip install mcp>=1.2.0
- Run:
- Error: "UV command not found"
- Install UV:
curl -LsSf https://astral.sh/uv/install.sh | sh
- Install UV:
- Configuration
- Error: "Failed to start MCP server"
- Verify config.json syntax
- Check path separators (use \ on Windows)
- Ensure UV is in your system PATH
Contributing
- Fork the repository
- Create feature branch
- Submit pull request
License
MIT
Support
Issues: GitHub Issues
関連サーバー
FRED Economic Data
Access economic data from the Federal Reserve Bank of St. Louis (FRED).
Databricks MCP Server by CData
A read-only MCP server for Databricks, enabling LLMs to query live data without writing SQL.
MCP for Neo4j
Connects to Neo4j graph databases with ability to use GDS functions ( when available), a read only mode , and set the sample size for schema detection
DuckDB Knowledge Graph Memory
An MCP memory server that uses a DuckDB backend for persistent knowledge graph storage.
MySQL MCP Server
A MySQL database server for AI assistants, enabling full CRUD operations, transaction management, and intelligent rollback.
Vestige MCP
Provides comprehensive DeFi analytics and data for the Algorand ecosystem through the Vestige API.
PostgreSQL
Provides read-only access to PostgreSQL databases, enabling schema inspection and query execution.
Materials Project MCP
Query the Materials Project database using the mp_api client. Requires an MP_API_KEY environment variable.
CData Salesforce MCP Server
A read-only MCP server by CData that allows LLMs to query live Salesforce data. Requires the CData JDBC Driver for Salesforce.
PostgreSQL MCP Server
Provides read-only access to PostgreSQL databases using a connection string.