mcp-fi
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-fiShow me the top 5 customers by revenue from my_postgres"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP-FI
Query any database in plain English using Claude.
MCP-FI is a Model Context Protocol (MCP) server that connects Claude to your data sources — SQLite, CSV, PostgreSQL, and MySQL. Instead of writing SQL, you just ask questions and Claude figures out the rest.
How It Works
You: "Which customers have the highest lifetime value?"
↓
Claude calls summarise_data → understands your schema
↓
Claude calls query_data → generates and runs the SQL
↓
Claude explains the results in plain EnglishNo SQL knowledge required.
Related MCP server: MCP Database Server
Features
Multiple data sources — SQLite, CSV, PostgreSQL, MySQL
Three tools exposed to Claude:
summarise_data— understand table structure before queryingquery_data— run SQL against any configured sourcecompare_data— diff two sources and find differences
Plain English queries — Claude writes the SQL for you
Local or deployed — run locally via Claude Desktop or deploy via SSE for others to use
Requirements
Python 3.12+
Claude Desktop (for local use)
Node.js (for MCP Inspector testing)
Installation
1. Clone the repo:
git clone https://github.com/yourusername/mcp-fi.git
cd mcp-fi2. Create and activate a virtual environment:
python -m venv venv
source venv/bin/activate # Mac/Linux
venv\Scripts\activate # Windows3. Install dependencies:
pip install -r requirements.txt4. Set up your config:
cp config.example.json config.jsonThen edit config.json with your actual data source paths and credentials (see Configuration below).
Configuration
Edit config.json to define your data sources. Each source needs a unique name that you'll use when asking Claude questions.
{
"sources": {
"my_sqlite_db": {
"type": "sqlite",
"path": "/absolute/path/to/your/database.sqlite"
},
"my_csv_data": {
"type": "csv",
"data_dir": "/absolute/path/to/your/csv/folder/"
},
"my_postgres": {
"type": "postgres",
"dsn": "postgresql://user:password@localhost/dbname"
},
"my_mysql": {
"type": "mysql",
"host": "localhost",
"port": 3306,
"database": "mydb",
"user": "root",
"password": "yourpassword"
}
}
}Important: Always use absolute paths — relative paths will not work when Claude Desktop launches the server.
You can define as many sources as you need. Mix and match types freely.
Connect to Claude Desktop
Open your Claude Desktop config file:
Mac:
open ~/Library/Application\ Support/Claude/claude_desktop_config.jsonWindows:
%APPDATA%\Claude\claude_desktop_config.jsonAdd the mcpServers block:
{
"mcpServers": {
"mcp-fi": {
"command": "/absolute/path/to/mcp-fi/venv/bin/python",
"args": [
"/absolute/path/to/mcp-fi/server.py"
],
"cwd": "/absolute/path/to/mcp-fi"
}
}
}Restart Claude Desktop. You should see a 🔨 hammer icon in the chat input — that confirms MCP-FI is connected.
Usage Examples
Once connected, just ask Claude questions in plain English:
Explore your data:
What tables are available in my_sqlite_db?Query data:
Show me the top 10 customers by total orders from my_postgresFilter and aggregate:
What is the average net worth by income level in my_csv_data?Compare sources:
Compare the users table between my_sqlite_db and my_postgres — are they in sync?Complex analysis:
Which products in my_mysql have been ordered more than 100 times
and what is their average rating?Claude will automatically call summarise_data to understand your schema, then query_data to fetch results, and finally explain everything back to you clearly.
Testing
Run the full test suite:
pytest tests/ -vRun a specific connector test:
pytest tests/test_sqlite_connector.py -v
pytest tests/test_csv_connector.py -v
pytest tests/test_postgres_connector.py -v
pytest tests/test_mysql_connector.py -vThe suite covers 116 tests across all connectors, tools, router, and server — all using isolated temporary databases so no real data is touched.
Test With MCP Inspector
Before connecting to Claude Desktop, you can test all tools interactively:
npx @modelcontextprotocol/inspector python server.pyOpen the URL shown in your terminal, click Connect, then Tools to see and run all three tools manually.
Deployment (SSE)
To deploy MCP-FI as a shared server others can connect to:
1. Run in SSE mode:
python server.py --sse --port 80002. Deploy to Railway:
Create a Procfile:
web: python server.py --sse --port $PORTPush to GitHub, connect the repo to Railway, and deploy. You'll get a public URL like:
https://mcp-fi-production.up.railway.app3. Others connect to your server by adding this to their claude_desktop_config.json:
{
"mcpServers": {
"mcp-fi": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp-fi-production.up.railway.app/sse"
]
}
}
}Project Structure
mcp-fi/
├── connectors/
│ ├── base_connector.py # Abstract contract all connectors implement
│ ├── sqlite_connector.py # SQLite support
│ ├── csv_connector.py # CSV support via DuckDB
│ ├── postgres_connector.py # PostgreSQL support
│ ├── mysql_connector.py # MySQL support
│ └── router.py # Routes source names to correct connector
├── tools/
│ ├── query_tool.py # Run SQL against any source
│ ├── summarise_tool.py # Describe table structure
│ └── compare_tool.py # Diff two sources
├── tests/ # 116 tests across all layers
├── data/ # Put your SQLite and CSV files here
├── server.py # Entry point — stdio and SSE transport
├── config.json # Your data sources (gitignored)
├── config.example.json # Template for config.json
└── requirements.txt # Python dependenciesAdding a New Data Source
Add your source to
config.jsonAsk Claude: "What tables are in my_new_source?"
That's it. No code changes needed.
Supported Data Sources
Type | Config key | Notes |
SQLite |
| Provide absolute |
CSV |
| Provide absolute |
PostgreSQL |
| Provide |
MySQL |
| Provide |
License
...
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- Flicense-qualityDmaintenanceProvides Claude Desktop with secure access to multiple database connections, allowing users to query MySQL, PostgreSQL, SQLite, and SQL Server databases directly through natural language.
- Alicense-qualityDmaintenanceEnables Claude to connect to and interact with SQLite, SQL Server, PostgreSQL, and MySQL databases through natural language. Supports executing queries, managing tables, exporting data, and storing business insights with authentication options including AWS IAM.706MIT
- Alicense-qualityCmaintenanceProvides database access capabilities to Claude with support for both SQLite and SQL Server databases. It enables users to manage schemas, execute SQL queries, and export results through natural language interaction.706MIT
- Alicense-qualityDmaintenanceEnables natural language interaction with local SQLite databases through Claude Desktop, translating plain English queries into SQL for data analysis and exploration.3MIT
Related MCP Connectors
Query PostgreSQL databases in plain English — LLM-generated, safety-validated SQL.
WHOOP recovery, strain, sleep and workouts in Claude via official WHOOP OAuth. Free, open source.
Talk to your live-events CRM (campaigns, analytics, paid ads, segments) in Claude and ChatGPT.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/toyetola/mcp-fi'
If you have feedback or need assistance with the MCP directory API, please join our Discord server