Docs

Search reference.

By name, by content, or both at once: typical speed per access method.

Access methods

Four ways in. One engine.

Interlinked Desktop

search bar

by name

12ms

inside files

15ms

ifile CLI

shell & scripts

by name

8ms

inside files

10ms

MCP API

Fastest for agents

AI agents & LLMs

by name

0.8ms

inside files

2.5ms

Direct pipe

your own client

by name

0.2ms

inside files

0.6ms

Typical latency in milliseconds. Scoped = a query pinned to a folder; whole machine = every drive at once.

The tools

Four tools. Every search.

iFileNames2,000 ns

Find files and folders by filename across every drive.

iFileText~5 ms

Search text inside files. Functions, TODOs, imports, errors.

iFileSearchParallel

Not sure where the match is? Both at once.

iFileKnowledgeOne path

Everything known about one file or folder: its machine-wide connections, plus the notes on it.

Example queries

budget.xlsxFind a spreadsheet by name

ext:pdf annual reportPDFs with both words in the name

/src ext:ts mod:7dTypeScript in /src, changed this week

type:video sort:sizeAll videos, largest first

TODO: fixEvery TODO across your codebase

Syntax

Every operator composes.

Search

keywords

All words must match

budget 2025

"quotes"

Hold spaces in a folder name

/"ATS Pack/Wallpapers"

||

Either term (OR)

readme || license

!folder

Skip everything inside a folder

config !test

Filter

/folder

Scope to directory

/src auth

ext:

By extension

ext:ts,tsx

type:

By category

type:video

size:

By file size

size:>10mb

mod:

Modified within

mod:7d

created:

Created within

created:30d

show:

Add back hidden and system files

show:all

Sort

sort:relevance

Best match (default)

sort:modified

Recently changed

sort:size

Largest first

sort:name

A → Z

sort:created

Newest created

File types

type:codetype:documenttype:videotype:audiotype:imagetype:spreadsheettype:presentationtype:applicationtype:repotype:other

Time and size sorts run newest and largest first: append -asc to flip one. sort:size-asc = smallest first. Name already runs A → Z, so it flips the other way: sort:name-desc = Z → A.

Verified performance

Every query kind, measured: medians, not cherry-picks.

Real end-to-end latency with full ranking on a live, freshly cold-built index: the honest distribution, worst cases and all. Beside each, what the same job costs the scanner your AI uses today.

465

timed queries

0

errors

Query kind	median	p95	worst	ripgrep, same job
File name, selective	0.4 ms	2.0 ms	2.8 ms	no filename mode, the OS walk: 88 sec
File name, common, full-set ranked	5.4 ms	9.1 ms	10.3 ms	no filename mode
Folder name	1.8 ms	3.4 ms	3.9 ms	no folder search
Content, common words	9.0 ms	12 ms	33 ms	re-reads every file, every query
Content, rare symbols	7.0 ms	60 ms	91 ms	17 min 13 sec (whole machine, cold)
Content, exact phrases	16 ms	167 ms	167 ms	re-scans every file: the more text, the slower
Document text, PDF · DOCX · XLSX	3.0 ms	7.0 ms	13 ms	can't: scanners don't read PDF/Office text
Scoped to a folder	0.2 ms	0.4 ms	0.4 ms	~90 ms (one warm folder, their best case)

File name, selective

0.4 msp95 2.0 msworst 2.8 ms

ripgrep: no filename mode, the OS walk: 88 sec

File name, common, full-set ranked

5.4 msp95 9.1 msworst 10.3 ms

ripgrep: no filename mode

Folder name

1.8 msp95 3.4 msworst 3.9 ms

ripgrep: no folder search

Content, common words

9.0 msp95 12 msworst 33 ms

ripgrep: re-reads every file, every query

Content, rare symbols

7.0 msp95 60 msworst 91 ms

ripgrep: 17 min 13 sec (whole machine, cold)

Content, exact phrases

16 msp95 167 msworst 167 ms

ripgrep: re-scans every file: the more text, the slower

Document text, PDF · DOCX · XLSX

3.0 msp95 7.0 msworst 13 ms

ripgrep: can't: scanners don't read PDF/Office text

Scoped to a folder

0.2 msp95 0.4 msworst 0.4 ms

ripgrep: ~90 ms (one warm folder, their best case)

Real searches on one real PC, a Ryzen 9 on Windows 11, same machine and same queries for both. And ripgrep isn’t a strawman: it’s the search engine inside VS Code and the one AI agents like Claude Code run. Full methodology in the ripgrep comparison.

One engine. Four ways in.

The desktop app, the CLI, the MCP server and the direct pipe: one install, one always-fresh index. Free on every machine you own.

Download for Windows