keboola/mcp-server
Author: keboola
Description: Model Context Protocol (MCP) Server that exposes Keboola project data, SQL transformations and job controls as callable tools for AI agents (Claude, Augment Code, LangChain, etc.).
Stars: 68
Forks: 14
License: MIT License
Category: Open Source
Overview
Installation
1. Ensure Python 3.9+ and Git are installed.
2. Clone the repository:
git clone https://github.com/keboola/mcp-server.git
cd mcp-server3. (Optional) create and activate a virtual-environment:
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate4. Install server and its dependencies:
pip install -U pip
pip install -r requirements.txt
# If a Makefile or setup.py is present you can also run:
# make install or pip install -e .5. Configure the MCP server before first run:
- Create .env (or export variables) supplying at minimum:
- Adjust any YAML/JSON config files found in the config/ directory to point to your Keboola project, buckets, and tables.
MCP_PORT=8080 # Listening port
MCP_STORAGE_URL=<keboola-storage-api-endpoint>
MCP_STORAGE_TOKEN=<keboola-storage-api-token>
LOG_LEVEL=INFO # DEBUG | INFO | WARNING | ERROR6. Start the server locally:
python -m mcp_server # or: uvicorn mcp_server.main:app --reload7. Verify the health endpoint:
curl http://localhost:8080/health8. (Optional) Build a production container image:
docker build -t keboola/mcp-server:latest .
docker run -d -p 8080:8080 --env-file .env keboola/mcp-server:latestLicense: MIT License
Updated 7/30/2025