Skip to content

OpenDataSci

OpenDataSci is a production-grade AI agent for data science and machine learning. Point it at a CSV, Parquet, Excel, or any other data file and ask questions in plain English — it writes Python, executes it in an isolated sandbox, and returns answers, charts, and models.

Key features

  • Conversational analysis — multi-turn sessions with full context; follow-up questions build on previous answers
  • Sandboxed code execution — all Python runs in an isolated sandbox; no side-effects to your environment
  • Concurrent workers — complex tasks (e.g. "compare five models") are split across concurrent sub-agents automatically
  • Every major LLM provider — Anthropic, OpenAI, AWS Bedrock, Google Gemini, Vertex AI, Azure OpenAI, Ollama, and any OpenAI-compatible server (e.g. vLLM)
  • Adaptive thinking — long-horizon reasoning via Anthropic and Bedrock's adaptive-thinking mode (Claude 4.6+)
  • Skills system — drop Markdown skill files into .opendatasci/skills/ to inject domain methodology
  • MCP tool servers — connect the agent to external tool servers via .opendatasci/mcp.json

Quick start

pip install open-data-sci
export ANTHROPIC_API_KEY=sk-ant-...
opendatasci data.csv

from opendatasci import create_agent

async with create_agent("data.csv") as agent:
    async for event in agent.astream("Summarise this dataset."):
        if event.type == "token":
            print(event.content, end="", flush=True)