ChatBI

The ChatBI MCP Server lets you run an MCP-compatible server locally or in your environment to perform automated data analysis, charting, and LLM-powered code generation. It exposes a controllable interface that clients can connect to for executing data tasks and generating results with configurable models. This guide shows you how to install, run, and use the server effectively.

How to use

Configure your MCP client to connect to the ChatBI MCP Server using a local stdio setup. You will run the server locally and connect the client to it. You can verify the server is ready by ensuring the configured tools appear in the client interface and that you can trigger data analysis workflows.

Prepare the environment and start using the server with a client like Cherry Studio or other MCP-capable tools. You will load your server configuration, then confirm that the tools list is populated. If tools list is visible, the configuration is correct.

For visualization, you can pair the server with charts through the mcp-server-chart integration to generate dashboards and dashboards-ready outputs.

Your workflow can automate data analysis steps, including generating a data analysis plan, executing the plan, producing results, and rendering a dashboard. Each step feeds into the next to create a cohesive analytical pipeline.

How to install

Prerequisites: you need Python 3.11 and a compatible shell to run commands. The server uses a Python-based MCP runtime, so ensure Python is installed on your system.

# Prepare Python virtual environment
uv venv .venv --python=3.11

# Activate environment
source .venv/bin/activate

# Install dependencies
uv pip install -r requirements.txt

Prepare the LLM configuration by copying the example environment and editing it with your preferred model. This helps ensure high-quality code generation for your tasks.

Start the server after installing dependencies and configuring the environment.

cd src
python pandas_mcp_server.py

Configuration and tips

Create and edit the environment file to set your LLM provider parameters. You can customize model choices to achieve better code generation quality, such as higher-capacity models when available. Keep your sensitive keys secure and avoid exposing them in logs.

In the client configuration, set a relatively generous timeout in case LLM code generation takes longer. If errors occur during generation, you may need to retry or adjust the LLM model selection.

The server is started from the src directory with the Python script pandas_mcp_server.py. The client should be configured to communicate with this local stdio endpoint.

Notes and troubleshooting

If the server does not start, ensure you activated the Python virtual environment and installed dependencies correctly. Verify you are running Python 3.11 compatible with the project.”

If the tools list does not populate in the client after configuration, double-check the configuration path, ensure the client has permission to access the local server, and retry starting the server.

Security and maintenance

Keep your LLM API keys secure. Do not commit .env files to public repositories. Rotate keys periodically and use environment-specific configurations to limit exposure.

Appendix: Example tool visualization setup

You can enable visualization using the mcp-server-chart integration. This lets you render dashboards based on the analyses performed by the MCP server, enhancing insights and sharing outcomes with your team.

Available tools

mcp-server-chart

Visualize server outputs with charts using the mcp-server-chart integration