RehanMohammed985/mcp-causal-inference-server
If you are the rightful owner of mcp-causal-inference-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.
The MCP Causal Inference Server is a Python-based server that performs causal inference analysis on synthetic customer spending data using DoWhy and FastMCP frameworks.
MCP Causal Inference Server
A Python-based MCP server implementing causal inference analysis on synthetic customer spending data. Uses DoWhy for causal effect estimation and FastMCP for exposing causal tools. Enables interactive causal analysis with treatment and outcome variables from program signup data.
This repository contains a Python MCP server (server.py) that performs causal inference analysis on synthetic customer spending data.
The project uses the FastMCP framework to expose causal analysis tools and DoWhy to estimate causal effects using methods like propensity score matching and linear regression.
Features
- Synthetic dataset simulating user spending before and after program signup
- Causal graph modeling treatment effects and confounders
- MCP tools for estimating causal effects and querying variable relationships
- Easily extensible for custom causal inference workflows
Setup
-
Create a Python virtual environment and activate it: '''bash python -m venv venv source venv/bin/activate # or venv\Scripts\activate on Windows
-
python -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows
-
python server.py
Example Dataset Variables
| Variable | Description |
|---|---|
treatment | Whether the user signed up for the program |
pre_spends | Spending before treatment month |
post_spends | Spending after treatment month |
signup_month | Month when the user enrolled |
Z | Confounder used in the causal graph |
💡 Project Purpose
This server demonstrates how language-based tool interfaces can support causal analysis workflows using natural language interfaces. The setup is useful for researchers, data scientists, or ML systems that need interpretable causal inference.
Tools Available
@mcp.tool() get_causal_estimate(treatment, outcome) Estimates causal effect using backdoor methods.
@mcp.tool() query_relationship(treatment, outcome) Checks if a causal relationship is identifiable using backdoor/frontdoor/IV.
@mcp.tool() get_variable_descriptions() Returns descriptions of the available dataset variables.