ig instant-grep
ig v2.3.0 vs rtk 0.42.2 — 2026-06-05

ig vs RTK

ig ships 141 TOML filter rules and 40 dedicated structured parsers — broader coverage than rtk's claimed "100+ commands", plus a trigram-indexed search engine rtk does not have. On a real benchmark, ig emits 15.5% fewer bytes at identical speed.

141
TOML filter rules across 40 files
40
Dedicated JSON/structured parsers
144K / 171K
ig vs rtk total bytes (21 cases, v2.3.0)
-84%
ig signature reads vs rtk (2.3K vs 14.8K)

Filter Coverage — ig vs RTK

Audited 2026-06-05. ig covers 90+ unique tools across 40 TOML filter files (141 rules) and 40 dedicated JSON/structured parsers. RTK claims "100+ commands" but documents ~63. ig has significantly more coverage in IaC, containers, lint, PHP, .NET, monorepo, and system tools.

Category ig rtk
Test runners 12 8
Lint / format 16 7
Build 7 2
Containers / K8s 23 7
Cloud / IaC 18 8
Package managers 8 4
Git platforms 3 1
System / network 13 4
Monorepo 2 0
.NET / Elixir / Swift / Java 4 0
Total 106 41

Count = distinct tool/subcommand entries. ig's total is verifiable: grep -c '\[\[filters\]\]' filters/*.toml → 141 rules. RTK's internals are not disclosed.

Real Benchmark — 21 Cases

Benchmark run on the instant-grep codebase itself (~90 Rust source files). ig v2.3.0 vs rtk 0.42.2, Linux x86_64, ripgrep 15.1.0. Each command timed with date +%s%N, output measured with wc -c.

Category Command ig bytes rtk bytes Note
search literal: fn main 2,858 3,128
search regex: impl.*Display 29 29
search count: pub fn 1,682 1 rtk -c unsupported
search files-only: todo 198 0 rtk -l unsupported
search --top 5: filter moat 2,553 16,088
search --semantic: error moat 2,291 16,384
read full: main.rs 80,476 69,130 ig keeps line-number prefix
read sigs: main.rs moat 2,354 14,833
read full: pipeline.rs 15,005 12,509 ig keeps line-number prefix
read sigs: pipeline.rs moat 1,849 3,989
smart smart: src/filter/ 36 0 rtk smart dir unsupported
smart smart: src/ 149 0 rtk smart dir unsupported
listing ls: root 314 286
listing ls: src/ 576 548
git git status 253 271
git git log 923 1,270
git git diff 6,662 5,233
tools env 3,346 1,026
tools deps 397 256
tools json: Cargo.toml 0 0
tools diff: 2 files 22,296 25,745

Aggregate Totals

ig total bytes: 144,247

rtk total bytes: 170,726

ig total time: 2,327 ms

rtk total time: 2,323 ms

Summary

ig emits 15.5% fewer bytes at identical speed (~110ms/cmd).

The moat rows (--top, --semantic, signatures) drive the aggregate advantage: 80-86% byte reduction on queries RTK cannot rank or expand.

Raw CSV: results.csv, summary.md. Previous benchmark (v1.10.0, 115 cases): results.csv, summary.md.

The Index Moat

rtk is a generic output-compression proxy: it shells to rg, git, find... and compresses their output. ig is a search engine with a persistent index — which unlocks capabilities rtk cannot match without completely redesigning itself.

--top N (BM25 ranking)

Score each matched file by Okapi BM25 and return only the top N. rtk has no tf / df / avdl so it cannot rank — it can only cap flat output length. 

ig --top 5 "filter"
  →  2,553 B

rtk grep "filter"
  →  16,088 B   (-84%)

Signature reads (-s)

ig extracts imports + function signatures. rtk strips bodies but keeps far more noise. The gap widens on large files. 

ig read src/main.rs -s
  →  2,354 B

rtk read src/main.rs -l aggressive
  →  14,833 B   (-84%)

RTK Functional Gaps

Verified on rtk 0.42.2 (2026-06-05). These are features ig supports that rtk does not.

  • grep -c (count per file) — rtk returns "0" regardless of actual match count.
  • grep -l (files-with-matches) — rtk returns empty output.
  • smart on directories — rtk fails with "Is a directory". ig supports directory-level summaries.
  • symbols, context, pack — ig-only capabilities with no rtk equivalent.

Where rtk Still Wins

  • Platform integrations (14 vs 3). rtk supports Cursor, GitHub Copilot, Windsurf, Cline, and other editors out of the box. ig currently targets Claude Code, Codex, and Gemini CLI.
  • Full file reads (~15% smaller). ig keeps the 42: content line-number prefix because it's what lets the Edit tool round-trip precisely. rtk drops it — fewer bytes, less precision.
  • env / deps compression. rtk's env filter is more aggressive (1.0K vs 3.3K). ig's filter prioritizes security (credential masking) over raw byte count.

Feature Comparison

Feature ig rtk
Trigram-indexed search engine moat Yes No (wraps rg)
BM25 --top N ranking moat Yes No (no index)
--semantic PMI query expansion moat Yes No (no index)
IDF-weighted sparse n-grams moat Yes No
Bloom + locMask (3.5-gram) moat Yes No
ig read -s / smart / symbols / pack moat Yes File-only smart
Budget mode (entropy-scored) moat Yes No
Relevance boost (-r) moat Yes No
Delta reads (git changes) moat Yes No
grep -c / -l / -i flags moat All supported Not supported
Directory-level smart summaries moat Yes File-only
TOML filter rules 141 rules Not disclosed
Dedicated structured parsers 40 parsers Not disclosed
Command-output filter wrappers 40+ wrappers 100+ claimed
0/1/2/3 hook exit-code protocol Yes Yes
Permission engine (deny / ask) Yes Yes
RTK filter import ig import-rtk —
Agent setup / hook install 11 agents 14 platforms
Token savings dashboard ig gain rtk gain
Git output compression Yes Yes
ls / tree compression Yes Yes
Process-per-invocation (no daemon) Yes Yes

moat rows mark structural advantages — they require a persistent index, which rtk does not have.

ig gain Dashboard

Real output from ig gain — token savings tracked across all sessions.

ig gain 
ig Token Savings
================================================================

Total commands:    3,445
Input bytes:       28.7M
Output bytes:      21.4M
Bytes saved:       7.3M (25.4%)

By Command  (top 20)
-----------------------------------------------------------------------
  #   Command                        Count    Saved    Avg%  Impact
-----------------------------------------------------------------------
  1.  ig read -s                       343     3.9M  92.7%  ==========
  2.  ig git diff                      184     1.9M  70.6%  =====
  3.  ig run curl                       30   716.0K  92.1%  ==
  4.  ig run cargo test                  43   555.9K  76.1%  =
  5.  ig git log                       145   476.6K  46.2%  =
  6.  ig ls                            491   358.1K  77.2%  =
  …
 20.  ig run /tmp/bench/mocks...        14    24.5K  29.3%
-----------------------------------------------------------------------
Showing top 20 of 143 commands. Use `ig gain --full` to see the full list.

Reproduce

All benchmark data is published alongside each release: