18-tool CSV data engine via csv-mcp — read, query (SELECT/WHERE/ORDER BY), aggregate (GROUP BY), merge (JOIN), search, transform, stats, chart data, formula, cell/row/column updates

csv-mcp v1.4.0 — an 18-tool MCP server for reading, querying, aggregating, merging, transforming, analyzing, and editing CSV files. Credential-free, stdio transport, npm install.

Install & launch

npm install --prefix /tmp/csv-mcp csv-mcp @modelcontextprotocol/sdk

Entry point: node node_modules/csv-mcp/dist/server.js — auto-starts (no subcommand needed).

MCP client connection (Node SDK)

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "node",
  args: ["node_modules/csv-mcp/dist/server.js"],
  cwd: "/tmp/csv-mcp"
});
const client = new Client({ name: "my-agent", version: "1.0.0" });
await client.connect(transport);
const { tools } = await client.listTools(); // → 18 tools

Tools (all 18)

readcsv, writecsv, querycsv, transformcsv, analyzecsv, searchcsv, mergecsv, aggregatecsv, getcsvinfo, appendcsv, applyformula, generatechart, calculatestats, updatecell, updaterow, updatecolumn, updatecellsconditional, deleterows

Verified traces (26/26 passed, 0 failures, p50 1ms, max 3ms)

Test data

10-row sales CSV (id, product, category, price, quantity, region, date, status) + 4-row regions CSV (region, manager, budget).

1. read_csv — 3ms

{"tool": "read_csv", "args": {"filepath": "/tmp/csv-mcp/test-sales.csv"}}

Returns {success: true, headers: [...8 cols], rowCount: 10, rows: [...]}.

2. read_csv with pagination — 1ms

{"tool": "read_csv", "args": {"filepath": "/tmp/csv-mcp/test-sales.csv", "limit": 3, "offset": 2}}

Returns 3 rows starting from index 2.

3. getcsvinfo — 1ms

Returns file metadata: {path, exists: true, size: 610, headers: [8], columnCount: 8, rowCount: 10}.

4. query_csv — SELECT + WHERE + ORDER BY — 1ms

{"tool": "query_csv", "args": {"filepath": "/tmp/csv-mcp/test-sales.csv", "select": ["product","price","quantity"], "where": {"column": "category", "operator": "eq", "value": "Electronics"}, "orderBy": [{"column": "price", "direction": "desc"}]}}

Returns 4 rows, sorted: Screen Pro (299.99), Screen Lite (199.99), Widget B (49.99), Widget A (29.99).

5. query_csv — WHERE gt + LIMIT — 0ms

Filter price > 50 returns 2 rows (Gadget X 99.95, Gadget Y 149.50 — string comparison).

6. search_csv — 1ms

{"tool": "search_csv", "args": {"filepath": "/tmp/csv-mcp/test-sales.csv", "searchTerm": "widget"}}

Case-insensitive search: returns 2 matches (Widget A, Widget B).

7. search_csv with column filter — 0ms

Search "discontinued" in status column → 2 rows.

8. analyze_csv (full) — 1ms

Returns per-column statistics: count, nullCount, uniqueCount, dataType, min, max, sum, mean, median, stdDev, topValues, mode.

9. analyze_csv (single column: price) — 0ms

{"column": "price", "count": 10, "min": 8.75, "max": 299.99, "sum": 895.64, "mean": 89.56, "median": 39.99, "stdDev": 93.15}

10. calculate_stats — 0ms

{"tool": "calculate_stats", "args": {"filepath": "/tmp/csv-mcp/test-sales.csv", "columns": ["price","quantity"], "operations": ["sum","mean","median","min","max","stddev","percentile"]}}

Returns: price p25=19.99, p50=49.99, p75=149.50, p90=199.99, p95=299.99. Quantity sum=2015, mean=201.5.

11. aggregate_csv — GROUP BY category — 1ms

{"tool": "aggregate_csv", "args": {"filepath": "/tmp/csv-mcp/test-sales.csv", "groupBy": ["category"], "aggregations": [{"column": "price", "func": "avg", "alias": "avg_price"}, {"column": "quantity", "func": "sum", "alias": "total_qty"}, {"column": "id", "func": "count", "alias": "product_count"}]}}

Returns: Electronics (avg 144.99, qty 320, count 4), Hardware (avg 67.68, qty 1320, count 4), Tools (avg 22.49, qty 375, count 2).

12. aggregate_csv — GROUP BY region — 0ms

Returns: North (max 299.99, qty 380), South (max 199.99, qty 315), East