S
Claude CodeLet an agent run cells and query your real database
AI Data Analyst: Jupyter + Postgres MCP Notebook Lab
Setuproll editorial@setuproll92.0Overall score
An agentic exploratory-analysis setup where the model writes notebook cells, executes them, reads the outputs, and pulls live tables through a Postgres MCP server. Built for analysts and data scientists who want a real read-eval loop instead of copy-pasting code that never runs.
92.0Score
5Components
Get this build
terminal
claude mcp add postgres -- npx -y @modelcontextprotocol/server-postgresWhat gets written
- CLAUDE.md
- .claude/agents/eda-explorer.md
- .claude/agents/feature-engineer.md
- .claude/agents/chart-reviewer.md
- .mcp.json
Components
Model
- Claude Opus 5
- Claude Sonnet 5 for cheaper runs
MCP servers
- jupyter
- postgres
- filesystem
- github
Subagents
- eda-explorer
- feature-engineer
- chart-reviewer
Stack
- JupyterLab
- pandas
- Polars
- DuckDB
- Plotly
How it works
- Postgres MCP lets the agent inspect schemas and pull samples
- Jupyter MCP runs each cell and feeds outputs back to the model
- eda-explorer profiles columns, finds nulls, plots distributions
- chart-reviewer rejects misleading axes before you trust a figure
Summary
An agentic exploratory-analysis setup where the model writes notebook cells, executes them, reads the outputs, and pulls live tables through a Postgres MCP server. Built for analysts and data scientists who want a real read-eval loop instead of copy-pasting code that never runs.
92.0 score
0 Reviews
Your rating
Sign in to post
Loading discussion...