Huge files.
Fast answers.
Confident edits.

High-performance large-file viewer and editor

Inspect large CSV, JSONL, Parquet and Arrow datasets locally. Find the record that matters, review a correction and keep your original files intact.

20 GB mock CSV · Pro full-file filter after opening: 0.36 s. See measured results →

Version 0.1.1 · Windows x64 · macOS Apple Silicon · Install guide →

Free Preview available · No account or card · Compare plans

Local data workflowsInspect / Query / Edit
RowCove desktop with generated mock record IDs, categories and labels
Product screenshot, mock data only
CSV · JSONL · Parquet · Arrow · SQLite · DuckDBExplore supported formats →
Find the problem soonerFast local reads and focused queries
Make a reviewed correctionTrace records; save a separate copy
Work across formatsCSV, JSONL, Parquet, Arrow and more
Keep control of your dataLocal processing and deliberate AI sharing

Built for the training dataset, research export or large file that costs you time. Spend less of your day waiting on data, and more of it on the work and people that matter.

Serious speed.
Measured on large data.

20 GB CSV · Synthetic (mock) data

Our full-file column contains filter was about 14× faster than EmEditor Professional in normal mode in this completed 20 GB mock CSV test.

3.48 s

Full-file access + first 1,000 records

Includes preparation and all 16 fields of the first records.

0.36 s

Full-file filter after opening

Column contains filter across the file. EmEditor normal mode: 5.06 s.

10.14 s

Open through complete numeric sort

From opening to the complete sorted result and first records, including preparation.

RowCove Pro · Three-run medians on Windows. Timings stop when data is returned, before screen paint. Results vary by device and file.

Compare desktop data tools →

Your dataset is large.
Your next step should be clear.

For people working at data scale

Training data, market history or a multi-gigabyte export: see the records that answer the question, with their source context intact.

Find the record
behind the training issue.

AI training & data engineering

Inspect nested conversations, labels and preference pairs in JSONL, Parquet and Arrow. Follow a finding to its source shard, then open an editable source to review a correction.

Investigate training data

Keep the exact value.
Follow the evidence.

Quantitative research

Filter large market-data exports and backtest results. Inspect complete IDs, timestamps and numeric values, and keep source records with the research finding.

Investigate market data

Locate it. Review it.
Save a separate copy.

Large-file analysis

Move from a multi-gigabyte CSV or JSONL file to the records that need attention. Stage supported corrections, reopen the saved copy and keep the original intact.

View and edit large files

Fix the next issue.
Keep the original.

From finding to correction

Check the full value, trace its source record and review the change before saving a separate copy. A native Rust engine handles the data work while the desktop keeps the next step visible.

One label corrected in 1,048,576 mock JSONL records. The complete separate-copy save took a median of 2.38 s; every output record was checked and the original remained unchanged.

Save only, after opening · Warm cache · Three Windows runs. This is a JSONL save measurement, separate from the 20 GB CSV benchmark.

Watch the JSONL correction
01Locate the recordChoose useful fields, filters and source partitions
02Review the correctionInspect complete values and stage supported changes
03Save a separate copyKeep the source intact; reopen the output to verify

CSV, quant, AI
A shorter path to the answer

Purpose-built data workflows

Choose a synthetic example to explore the workflow. The illustration shows field selection—not measured speed or guaranteed byte savings.

Mock data only

Inspect the complete value, narrow the source query, save a reviewed copy.

Project only useful fields
Source fields
24
Requested fields
4

Synthetic field counts. Not a performance benchmark.

Synthetic result previewCSV workflow guide →
Generated mock records; not customer or research data
record_idcategoryamountstatus
mock_001Demo A125.00review
mock_002Demo B98.50ready
mock_003Demo A210.25review

CSV synthetic example selected

Work with the data
you already produce

One workspace, many formats

Text exports, training records, columnar shards and local databases—with the reader and available operations made explicit.

Text & records
CSV · TSV · DSV · TXT · LOG · JSON · JSONL · NDJSONInspect tables and nested records; use supported query and copy workflows.
Columnar & spreadsheets
Parquet · Arrow IPC · Feather · XLSX · XLSTyped fields and bounded pages; capabilities depend on the reader.
Collections & databases
JSONL / Parquet / Arrow shards · SQLite · DuckDBChoose partitions, projected fields or an explicit read-only database table.
Compressed & specialist
GZIP · Zstandard · BZIP2 · ZIP · TAR · NPY / NPZ · Zarr · Safetensors · TFRecordCompressed preparation, archive catalogs or scientific metadata/framing inspection as supported. These do not all decode samples or allow edits.
Full format and operation guide →

Your dataset stays
on your machine

Local data, under your control

Parsing, indexing, queries and supported edits run locally in the Rust data engine. Dataset upload is not required to inspect your files.

Account activation and updates use network services. MCP sends the requested results to the AI client you choose; a cloud-backed client may pass them to its model provider.

Understand the privacy boundary →
Local data-processing path
  1. Your source filesOn this device
  2. Native Rust engineParse · index · query
  3. Your desktop resultsInspect · review · save a copy

MCP sharing is a separate, explicit connection.

Finish the data task
Keep the rest of your day

The point of faster data work

Fewer temporary scripts. Less manual file splitting. Less staring at a frozen table.

Spend your attention on the finding—and leave room for dinner with family, a walk outside or a game with friends.

Find the workflow that saves you steps →
A focused data request
Inspect the training-data schema.
Choose one split and partition.
Return only the required fields.
Include the source file and row.
Do not edit or export yet.
Tool responses go to the AI client you connect.

Ask your AI.
Inspect the evidence.

AI-assisted data work

Let your connected assistant inspect schemas and query local data through MCP. Bring back useful records with their source context, then review the evidence in RowCove. Test tools and inspect call history in the desktop client.

Connect your AI tools

Choose the workflow
that saves you time.

A tool that earns back your time

Try your own file with Preview. Pro adds supported full-file queries, quality checks, editing, export and local MCP workflows.

Get to know your data.

Preview

$0 / forever

No account. No card. Start with your own file.

  • Supported source discovery and schema
  • Read-only first page, up to 100 records
  • Inspect nested values and field types
  • Your original file stays untouched

Before you open your next large file

Questions, answered

Contact support →
Who is RowCove designed for?

RowCove is a desktop data workspace for people whose files have outgrown a comfortable spreadsheet workflow: AI data engineers inspecting pretraining or post-training corpora, quantitative researchers working with market history and backtest exports, and analysts opening multi-gigabyte CSV or columnar files.

Use it to understand the schema, find relevant records, inspect complete values and save reviewed changes as a separate copy. It is especially useful when you need to answer a focused question about a large local source without first uploading it or writing an import script.

Explore workflows for your data

How does RowCove work with files larger than a spreadsheet can comfortably handle?

The native Rust engine handles file access and data operations locally. The desktop displays bounded pages and renders visible records, so browsing does not require turning the entire source into a browser-sized table. Supported readers can also select relevant columns or partitions before returning records.

The operation still matters: a first page, a complete filter, a global sort and a saved copy have different preparation, memory and disk costs. Start with Preview to check your file and its reader, then use the supported Pro workflow for the full-file task you actually need.

Understand large-file performance Check reader capabilities

What can I check before buying Pro?

Preview is free and does not require an account or payment card. Open a supported file or dataset folder, review the detected source and available operations, and inspect a read-only first page of up to 100 records. Use the record inspector to check complete cell values and nested structures where the reader supports them.

Preview lets you check whether your source opens and whether the viewing workflow suits you. It does not provide a complete file count, full-file query, edit/export workflow or local MCP access. Those Pro capabilities remain reader-dependent; a recognized extension alone does not mean every operation is available.

Download and try your file Compare Preview and Pro

Do my datasets ever have to leave my computer?

You do not need to upload a dataset to RowCove to inspect, query or edit it. Those desktop operations run on your device. Account activation and billing use separate online services for information such as your purchase email, subscription and device access; opening a dataset does not send its contents to those services.

MCP is an explicit sharing path: requested tool results go to the AI client you connect. If that client uses a cloud model, results may reach its provider. Choose the folders, fields and result size you intend to share, and review the client’s own privacy policy before using sensitive data.

Read the privacy policy Control MCP sharing

Will editing overwrite my original dataset?

Sources open read-only by default. Where the reader supports editing, switch to Edit and stage your changes. Save copy writes the result to a separate destination, keeping the original available for comparison and recovery.

Before saving a large result, check the destination and free disk space. Open the new copy and verify representative records, types and the intended changes. Some readers are read-only or expose metadata rather than editable records; use the operations shown for the detected source.

Follow the save-copy workflow Check editing support

Which Windows and Mac computers are supported?

RowCove 0.1.1 is available as a Windows x64 installer for Windows 10 or later and as a macOS DMG for Apple Silicon Macs running macOS 12 or later. The Mac download is for M-series chips; this release does not include an Intel Mac build.

Choose the package for your computer on the download page. On Windows, run the installer; on macOS, open the DMG and drag RowCove into Applications. If installation is blocked, contact support with the exact system message and your operating-system version.

Choose your download Get installation help

Get the answer.
Get your time back.

Windows and macOS

Bring your next data question to a fast local workspace. Inspect the record, review the next step and keep your files on your machine.

Version 0.1.1 · Windows x64 · macOS Apple Silicon · Install guide →