Privacy-First · Local · No Upload

The local data runtime
for AI agents.

Give Claude Code, Cursor and other AI agents safe, queryable access to CSV, Excel, Parquet and JSON files — without uploading your data.

Download Free → curl -fsSL https://pipetable.com/install | sh
— Real CLI session
pipetable ~/data/
$ pipetable ~/data/ Found 3 files (3×CSV) ✓ sales 5 cols 1.2MB 12ms new ✓ customers 4 cols 890KB 8ms new ✓ orders 6 cols 4.8MB 18ms new Done. 3 new Model: qwen2.5-coder:1.5b
> show me top 5 customers by revenue Using: customers, sales Thinking..... SELECT c.name, SUM(s.revenue) AS total FROM customers c JOIN sales s ON s.customer_id = c.id GROUP BY c.name ORDER BY total DESC LIMIT 5
5 row(s)
name total ──────────────────────────── Meridian Ltd 1284000 Vantage Group 892400 Apex Solutions 441200 Peak Systems 318900 BlueLine Inc 124600
> SELECT region, COUNT(*) AS deals FROM sales GROUP BY 1 ORDER BY 2 DESC
3 row(s)
region deals ─────────────── EU 24 US 18 APAC 9
The problem

AI agents are brilliant.
Until they need real data.

Today's AI coding assistants can generate code, write SQL and automate workflows. But when they need your local business data, developers usually have only three options: upload sensitive files, give unrestricted filesystem access, or copy-paste data into prompts. None of these scale.

Pipetable is the controlled middle ground — real SQL access, nothing uploaded.

Built for modern AI workflows
Private by default
Your files never leave your computer.
Real SQL, not hallucinations
AI agents execute real SQL queries instead of guessing answers.
Works with your favorite tools
Claude Code, Cursor, GitHub Copilot, Roo Code.
Lightweight
Native binary. Under 5 MB. Starts instantly.
Jobs people need done
Finance
Find duplicate invoices from the last two years.
Operations
Compare inventory exports across warehouses.
Sales
Summarize quarterly revenue by customer.
Engineering
Explore logs and CSV exports using SQL.
Data Analysis
Join multiple CSV files without importing them into a database.
Support
Search ticket exports for recurring issues.
MCP Server 01

Works inside
your AI coding tool.

Pipetable connects directly to Claude Code, Cursor and every MCP-compatible assistant, letting AI agents query your local data safely. A 5 MB binary. Nothing leaves your machine.

// .roo/mcp.json
{
  "mcpServers": {
    "pipetable": {
      "command": "pipetable",
      "args": ["mcp"]
    }
  }
}
// ~/.cursor/mcp.json
{
  "mcpServers": {
    "pipetable": {
      "command": "pipetable",
      "args": ["mcp"]
    }
  }
}
# one command, done
claude mcp add pipetable pipetable mcp
// .vscode/mcp.json
{
  "servers": {
    "pipetable": {
      "type": "stdio",
      "command": "pipetable",
      "args": ["mcp"]
    }
  }
}
Download Free → cargo install pipetable requires Rust macOS Universal arm64 + x86_64 · .tar.gz Linux x86_64 glibc · .tar.gz Linux ARM64 glibc · .tar.gz Windows x86_64 · .zip

MIT licensed — free for everyone. Source on GitHub →

On the roadmap
Today it's your terminal.
Tomorrow, your whole team.
Coming soon
Secure Workspaces
Define exactly what an AI agent can access.
Read-only folders
Permission policies
Query history
Shared workspaces
Audit logs
Built for developers
Free. Forever. No catches.
View on GitHub →
Use it, fork it, ship it. No license check, no usage limits, no commercial restriction. Questions? hello@pipetable.com
Native Rust MIT Licensed Open Source Cross-platform SQL-first MCP-native
Common questions
How do I give Claude Code access to my local data? +

One command: claude mcp add pipetable pipetable mcp. Then ask Claude to scan your folder and query your files. Pipetable gives Claude real SQL access via DuckDB — not guesses.

Can I query Excel files without uploading them? +

Yes. Pipetable reads .xlsx, .xls and .xlsm files directly from disk. No upload, no cloud, no size limit. Run pipetable ./yourfolder/ and your Excel files are immediately queryable.

Does it work offline, without an API key? +

Yes. SQL always works with no API key, no internet. For natural language, set ANTHROPIC_API_KEY or OPENAI_API_KEY, or run Ollama locally — no internet required.

How is this different from uploading to ChatGPT? +

Your data never leaves your machine. No size limit (ChatGPT truncates large files). Results are real DuckDB query output — ground truth, not generated. Sensitive data stays private by architecture, not just by policy.

What file formats are supported? +

CSV, TSV, Parquet, JSON, NDJSON, Excel (xlsx, xls, xlsm). Files up to 2GB. Folders scanned 3 levels deep. Hidden files and noise directories (node_modules, .git, target) skipped automatically.

Guides
GUIDE
Give Claude Code access to your local data
MCP setup, scan a folder, query with AI in 2 minutes.
GUIDE
Query CSV files without uploading to the cloud
SQL and natural language on local files. Any size.
GUIDE
Query Excel files with AI — no upload required
Read .xlsx directly from disk, query with DuckDB.
GUIDE
Natural language SQL with Ollama and DuckDB
Fully local NL queries. No API key needed.

AI agents need access to
real data.

Pipetable gives them a safe way to work with yours.

Download Free →
Stay in the loop

Get updates on new features

We'll let you know when new formats, performance improvements, or integrations ship. No spam, just what matters.

Unsubscribe anytime. We respect your privacy.