Process and Transform CSV Data

CSV Editor - AI-Powered CSV Processing via MCP

Stateful CSV editing for AI assistants. CSV Editor is an MCP server that gives Claude, ChatGPT, Cursor, Windsurf, and other MCP clients a full suite of CSV operations - with sessions, undo/redo, and auto-save built in. Most data MCPs are analyze-only; this one lets the AI edit.

🆕 What's new in v2.0.0 (April 2026)

• FastMCP 3.x - migrated from FastMCP 2 to 3.2, aligning with MCP spec 2025-11-25.
• Python 3.11+ required (was 3.10+). Tested against 3.11 / 3.12 / 3.13 / 3.14.
• --transport sse removed. Use --transport http (Streamable HTTP) for remote deployments. SSE was deprecated by FastMCP 3.
• Dependency refresh: pydantic 2.13, pyarrow 23, httpx 0.28.
• New CSV_EDITOR_CSV_HISTORY_DIR env var for configuring the history directory.
• First-class CI test matrix on GitHub Actions.

Users who pinned csv-editor>=1,<2 are unaffected and will continue to receive 1.x patches if needed. See CHANGELOG.md for the full list of breaking changes.

🎯 Why CSV Editor?

The Problem

AI assistants struggle with complex data operations - they can read files but lack tools for filtering, transforming, analyzing, and validating CSV data efficiently.

The Solution

CSV Editor bridges this gap by providing AI assistants with 39 specialized tools for CSV operations, turning them into powerful data analysts that can:

• Clean messy datasets in seconds
• Perform complex statistical analysis
• Validate data quality automatically
• Transform data with natural language commands
• Track all changes with undo/redo capabilities

Key differentiators vs. other CSV / tabular MCPs

When to pick CSV Editor: you want the AI to make changes to a CSV and iterate, not just answer questions about it. If your workload is read-only analytics on multi-GB files, a DuckDB-based MCP is likely a better fit; CSV Editor's DuckDB/Polars engine support is tracked on the roadmap.

⚡ Quick Demo

# Your AI assistant can now do this:
"Load the sales data and remove duplicates"
"Filter for Q4 2024 transactions over $10,000"  
"Calculate correlation between price and quantity"
"Fill missing values with the median"
"Export as Excel with the analysis"

# All with automatic history tracking and undo capability!

🚀 Quick Start (2 minutes)

Installing via Smithery

To install csv-editor for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @santoshray02/csv-editor --client claude

Fastest Installation (Recommended)

# Install uv if needed (one-time setup)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Clone and run
git clone https://github.com/santoshray02/csv-editor.git
cd csv-editor
uv sync
uv run csv-editor

Configure Your AI Assistant

{
  "mcpServers": {
    "csv-editor": {
      "command": "uv",
      "args": ["tool", "run", "csv-editor"],
      "env": {
        "CSV_MAX_FILE_SIZE": "1073741824"
      }
    }
  }
}

💡 Real-World Use Cases

📊 Data Analyst Workflow

# Morning: Load yesterday's data
session = load_csv("daily_sales.csv")

# Clean: Remove duplicates and fix types
remove_duplicates(session_id)
change_column_type("date", "datetime")
fill_missing_values(strategy="median", columns=["revenue"])

# Analyze: Get insights
get_statistics(columns=["revenue", "quantity"])
detect_outliers(method="iqr", threshold=1.5)
get_correlation_matrix(min_correlation=0.5)

# Report: Export cleaned data
export_csv(format="excel", file_path="clean_sales.xlsx")

🏭 ETL Pipeline

# Extract from multiple sources
load_csv_from_url("https://api.example.com/data.csv")

# Transform with complex operations
filter_rows(conditions=[
    {"column": "status", "operator": "==", "value": "active"},
    {"column": "amount", "operator": ">", "value": 1000}
])
add_column(name="quarter", formula="Q{(month-1)//3 + 1}")
group_by_aggregate(group_by=["quarter"], aggregations={
    "amount": ["sum", "mean"],
    "customer_id": "count"
})

# Load to different formats
export_csv(format="parquet")  # For data warehouse
export_csv(format="json")     # For API

🔍 Data Quality Assurance

# Validate incoming data
validate_schema(schema={
    "customer_id": {"type": "integer", "required": True},
    "email": {"type": "string", "pattern": r"^[^@]+@[^@]+\.[^@]+$"},
    "age": {"type": "integer", "min": 0, "max": 120}
})

# Quality scoring
quality_report = check_data_quality()
# Returns: overall_score, missing_data%, duplicates, outliers

# Anomaly detection
anomalies = find_anomalies(methods=["statistical", "pattern"])

🎨 Core Features

Data Operations

• Load & Export: CSV, JSON, Excel, Parquet, HTML, Markdown
• Transform: Filter, sort, group, pivot, join
• Clean: Remove duplicates, handle missing values, fix types
• Calculate: Add computed columns, aggregations

Analysis Tools

• Statistics: Descriptive stats, correlations, distributions
• Outliers: IQR, Z-score, custom thresholds
• Profiling: Complete data quality reports
• Validation: Schema checking, quality scoring

Productivity Features

• Auto-Save: Never lose work with configurable strategies
• History: Full undo/redo with operation tracking
• Sessions: Multi-user support with isolation
• Performance: Stream processing for large files

📚 Available Tools

⚙️ Configuration

Environment variables

Auto-Save Strategies

CSV Editor automatically saves your work with configurable strategies:

• Overwrite (default) - Update original file
• Backup - Create timestamped backups
• Versioned - Maintain version history
• Custom - Save to specified location

# Configure auto-save
configure_auto_save(
    strategy="backup",
    backup_dir="/backups",
    max_backups=10
)

🛠️ Advanced Installation Options

git clone https://github.com/santoshray02/csv-editor.git
cd csv-editor
pip install -e .

pipx install git+https://github.com/santoshray02/csv-editor.git

pip install csv-editor            # latest
pip install csv-editor==2.0.0     # pinned
# Or with uv:
uv tool install csv-editor

# Latest main
pip install git+https://github.com/santoshray02/csv-editor.git

# Specific release
pip install git+https://github.com/santoshray02/csv-editor.git@v2.0.0

# Or with uv
uv pip install git+https://github.com/santoshray02/csv-editor.git@v2.0.0

🧪 Development

Running tests

uv run pytest tests/ -v                  # Run tests
uv run pytest tests/ --cov=src/csv_editor # With coverage
uv run ruff check src/ tests/             # Lint
uv run black --check src/ tests/          # Format check
uv run mypy src/                          # Type check

CI runs the full pytest matrix on Python 3.11-3.14 for every push to main - see .github/workflows/test.yml.

Project Structure

csv-editor/
├── src/csv_editor/   # Core implementation
│   ├── tools/        # MCP tool implementations
│   ├── models/       # Data models
│   └── server.py     # MCP server
├── tests/            # Test suite
├── examples/         # Usage examples
└── docs/            # Documentation

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Quick Contribution Guide

1. Fork the repository
2. Create a feature branch
3. Make your changes with tests
4. Run uv run pytest tests/ and uv run ruff check src/ tests/
5. Submit a pull request

📈 Roadmap

Post-v2.0.0 priorities (see the 2026 relevance audit for context):

• pandas 3.0 / numpy 2.4 - Copy-on-Write migration, Arrow-backed default strings (follow-up to v2.0.0).
• DuckDB + Polars engines - swappable backends with DuckDB as the default for files >100 MB (closes the large-file gap).
• MCP async Tasks + Resource Links - non-blocking load_csv / export_csv / profile_data for GB files; paginated large results.
• Remote HTTP + OAuth (CIMD) - enables ChatGPT Connectors and VS Code Copilot remote usage.
• Elicitation - prompt for ambiguous CSV dialect / encoding / dtype at load time instead of failing.
• Docs migration - Docusaurus → MkDocs-Material with mkdocstrings for auto-generated API docs.

💬 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Documentation: Wiki

📄 License

MIT License - see LICENSE file

🙏 Acknowledgments

Built with:

• FastMCP - Fast Model Context Protocol
• Pandas - Data manipulation
• NumPy - Numerical computing

Ready to supercharge your AI's data capabilities? Get started in 2 minutes →