Backed by mandō

Extract fields.
At line speed.

Slice is the precision field extraction engine that cuts structured data into exactly the columns you need. Delimiter-aware, byte-precise, character-perfect — the extraction primitive every data pipeline secretly depends on.

See it live
slice
$ echo "name:email:role" | slice -d':' -f2
email
$ slice -d',' -f1,3 users.csv
alice,admin
bob,editor
carol,viewer
$ slice -c1-8 access.log
10.0.0.1
10.0.0.2
192.168.
$

Shipped by teams at

Vercel
GitHub
Cloudflare
Linear
Stripe
Docker
Datadog
CircleCI

Every extraction primitive.
One composable engine.

We took the most precise column extraction tool ever built and gave it a platform. Seven core features. Zero wasted bytes.

📊

FieldSelect™ -f

Select exactly the fields you need by position. Single fields, ranges, comma-separated lists — FieldSelect™ extracts columns from delimited data with zero ambiguity. The extraction primitive that makes CSV, TSV, and log parsing trivial.

🔗

DelimiterEngine™ -d

Specify any single character as your field separator. Tab by default, but DelimiterEngine™ handles colons, commas, pipes, semicolons — any character that separates your structured data into extractable columns.

✂️

CharSlice™ -c

Extract by character position. Fixed-width fields, positional data formats, log file prefixes — CharSlice™ cuts at exact character boundaries regardless of delimiters. Precision extraction for structured and unstructured data alike.

🧬

ByteCut™ -b

Select by byte offset for binary-safe extraction. Multi-byte character awareness, binary protocol parsing, raw data slicing — ByteCut™ operates at the lowest level of text extraction. When characters aren't enough.

🔄

Complement™ --complement

Invert your selection. Instead of specifying what to keep, specify what to drop. Complement™ flips the extraction set — select everything except fields 3 and 5. The inverse operation that makes exclusion as easy as inclusion.

🎯

StrictMode™ -s

Suppress lines that don't contain the delimiter. StrictMode™ filters out headers, comments, and malformed rows automatically — only lines with actual delimited structure make it through the pipeline.

OutputDelim™ --output-delimiter

Transform the delimiter on output. Read CSV, emit TSV. Parse colon-separated, output pipe-delimited. OutputDelim™ decouples input format from output format — the delimiter transformation primitive for format migration.

Read. Split. Extract. Emit.

01

Ingest the stream

Pipe any structured input — CSV files, log output, passwd entries, command results — into the Slice engine. Each line is processed independently, ready for extraction.

02

Split on delimiter

Slice tokenizes each line by your chosen delimiter. Tab by default, or any character you specify with -d. Every field gets a position number, starting at 1.

03

Select your columns

Specify fields by number, range, or list. -f1,3 for specific columns. -f2-5 for ranges. -c1-16 for character positions. -b1-8 for byte offsets. Precise extraction, every time.

04

Emit clean output

Selected fields stream to stdout, joined by the output delimiter. Pipe to the next tool, redirect to a file, or feed directly into your data pipeline. Zero overhead. Zero ambiguity.

0
Fields extracted per day
0
Avg. extraction latency
0
Delimiter detection accuracy
0
Supported delimiters

Scale your extractions.

From quick field grabs to enterprise-grade column extraction. Pick the tier that matches your data velocity.

Starter

Free

For individual developers parsing logs and CSVs.

  • Single field selection (-f)
  • Tab delimiter (default)
  • Character extraction (-c)
  • Up to 10K lines/invocation
  • Community support

Team

$79/mo

For engineering orgs shipping data at scale.

  • Everything in Pro
  • Complement™ (--complement)
  • OutputDelim™ (--output-delimiter)
  • Null-terminated mode (-z)
  • Unlimited throughput
  • Priority support

Enterprise

Custom

For organizations with compliance and SLA needs.

  • Everything in Team
  • Multi-file batch extraction
  • Custom output formatting
  • SSO & audit logging
  • 99.99% uptime SLA
  • Dedicated support engineer

Trusted by engineers who parse.

"We were writing Python scripts to extract columns from CSVs. Then we discovered Slice. One invocation of -d',' -f2,5,8 replaced 40 lines of pandas. We literally deleted a microservice."

J
Jordan Arevalo
Data Engineer, Segment

"The --complement flag is what sold me. Instead of selecting 18 out of 20 columns, I just exclude the 2 I don't need. Complement™ is the kind of feature that makes you rethink your entire ETL pipeline."

R
Raquel Mendes
Staff SRE, PlanetScale

"We parse 2TB of access logs daily. Slice with -d' ' -f1,4,7 extracts IP, timestamp, and path in under 90 seconds. Nothing else comes close. This is the data primitive we were missing."

T
Tariq Hassan
Platform Lead, Grafana Labs

"OutputDelim™ is underrated. We ingest colon-separated LDAP data and emit tab-separated output for downstream tools. Format conversion at extraction time. No intermediate transforms. Slice just gets it."

E
Elena Voronova
DevOps Engineer, Tailscale

"I showed StrictMode™ to my team and three people immediately refactored their shell scripts. Suppressing lines without delimiters should have been the default all along. Slice is the discipline we needed."

N
Nico Brandt
Engineering Manager, Render

See Slice extract in real time.

Real commands, real output. Click an extraction mode to see it execute.

slice demo
INPUT
OUTPUT

Why teams switch to Slice.

Capability Slice awk '{print $N}' Python / pandas
Field extraction ✓ Native, zero-config Requires format string Import + load + index
Delimiter flexibility ✓ Any single char ✓ -F flag sep= parameter
Byte-level precision ✓ -b flag Manual slicing
Complement / inversion ✓ --complement Manual negation .drop() method
Pipeline composability ✓ Native stdin/stdout ✓ Native Requires wrapper
Startup time ✓ <1ms ~2ms ~400ms

What's new in Slice.

v3.4

Null-terminated mode

-z / --zero-terminated uses NUL instead of newline as the line delimiter. Process filenames with embedded newlines and binary-safe data streams.

v3.3

Output delimiter override

--output-delimiter decouples input format from output format. Read comma-separated, emit tab-separated. The format migration primitive.

v3.2

Complement selection

--complement inverts your field/byte/character selection. Exclude instead of include — the inversion primitive for when specifying what to drop is easier than what to keep.

Answers to real questions.

-f selects by field position (delimiter-split columns). -c selects by character position. -b selects by byte offset. Use -f for structured delimited data like CSV/TSV. Use -c for fixed-width formats. Use -b when you need byte-level precision with multi-byte character sets. You can only use one mode per invocation.

Slice operates on single-character delimiters and does not parse quoted fields natively. For simple CSVs without embedded commas in quoted strings, -d',' -f works perfectly. For complex CSVs with quoted fields, pair Slice with a pre-processing step or use it downstream after normalization. The Unix philosophy: do one thing perfectly.

By default, Slice prints these lines unmodified — they pass through as-is. Use -s (StrictMode™) to suppress them entirely, ensuring only properly delimited lines appear in your output. This is critical for filtering headers, comments, and malformed rows.

Ranges are flexible: N selects one field, N-M selects an inclusive range, N- selects from N to end of line, -M selects from first to M. Combine with commas: -f1,3-5,8- selects field 1, fields 3 through 5, and field 8 onward. Selected input is always written in the same order it appears in the source.

Yes. Use --output-delimiter to specify a different delimiter for the output. For example, -d':' -f1,3 --output-delimiter=',' reads colon-separated input and emits comma-separated output. The default output delimiter matches the input delimiter.

Start extracting.

Join 9,000+ engineers who rely on Slice for every pipeline, every parse, every extraction.

No credit card. No sales call. Just precision field extraction.

See mandō's portfolio

The accelerator behind the tools that run the internet.