Token Savings

RTK saves 60-90% of tokens on common development operations by filtering command output before it reaches your LLM. This translates to lower costs, faster responses, and fewer context limit issues.

30-Minute Claude Code Session Breakdown

Real-world savings: A typical 30-minute coding session uses ~150,000 tokens without RTK. With RTK, this drops to ~45,000 tokens—a 70% reduction.

| Operation | Frequency | Standard | rtk | Savings | |-----------|-----------|----------|-----|---------|| | ls / tree | 10× | 2,000 | 400 | -80% | | cat / read | 20× | 40,000 | 12,000 | -70% | | grep / rg | 8× | 16,000 | 3,200 | -80% | | git status | 10× | 3,000 | 600 | -80% | | git diff | 5× | 10,000 | 2,500 | -75% | | git log | 5× | 2,500 | 500 | -80% | | git add/commit/push | 8× | 1,600 | 120 | -92% | | npm test / cargo test | 5× | 25,000 | 2,500 | -90% | | ruff check | 3× | 3,000 | 600 | -80% | | pytest | 4× | 8,000 | 800 | -90% | | go test | 3× | 6,000 | 600 | -90% | | docker ps | 3× | 900 | 180 | -80% | | Total | | ~118,000 | ~23,900 | -80% |

Estimates based on medium-sized TypeScript/Rust projects. Actual savings vary by project size and command usage patterns.

Real Examples: Before & After

Directory Listing

Standard ls -la (45 lines, ~800 tokens)

drwxr-xr-x  15 user  staff    480 Jan 23 10:00 .
drwxr-xr-x   5 user  staff    160 Jan 23 09:00 ..
-rw-r--r--   1 user  staff   1234 Jan 23 10:00 Cargo.toml
-rw-r--r--   1 user  staff   5678 Jan 23 10:00 Cargo.lock
drwxr-xr-x   8 user  staff    256 Jan 23 10:00 src
drwxr-xr-x   3 user  staff     96 Jan 23 10:00 tests
-rw-r--r--   1 user  staff   2345 Jan 23 10:00 README.md
-rw-r--r--   1 user  staff    567 Jan 23 10:00 .gitignore
-rw-r--r--   1 user  staff    123 Jan 23 10:00 LICENSE
...

RTK ls (12 lines, ~150 tokens)

📁 my-project/
├── src/ (8 files)
│   ├── main.rs
│   ├── lib.rs
│   └── filter.rs
├── tests/ (3 files)
├── Cargo.toml
├── Cargo.lock
├── README.md
└── .gitignore

Token savings: 81% (650 tokens saved)

Git Operations

Standard git push (15 lines, ~200 tokens)

Enumerating objects: 5, done.
Counting objects: 100% (5/5), done.
Delta compression using up to 8 threads
Compressing objects: 100% (3/3), done.
Writing objects: 100% (3/3), 342 bytes | 342.00 KiB/s, done.
Total 3 (delta 2), reused 0 (delta 0)
remote: Resolving deltas: 100% (2/2), completed with 2 local objects.
To github.com:user/repo.git
   abc1234..def5678  main -> main

RTK git push (1 line, ~10 tokens)

ok ✓ main

Token savings: 95% (190 tokens saved)

Test Output

Standard cargo test (200+ lines on failure)

Compiling my-project v0.1.0 (/Users/user/my-project)
    Finished test [unoptimized + debuginfo] target(s) in 2.34s
     Running unittests src/lib.rs (target/debug/deps/my_project-abc123)

running 15 tests
test utils::test_parse ... ok
test utils::test_format ... ok
test utils::test_validate ... ok
test core::test_init ... ok
test core::test_process ... ok
test core::test_filter ... ok
test filters::test_minimal ... ok
test filters::test_aggressive ... ok
test tracking::test_insert ... ok
test tracking::test_query ... ok
test git::test_status ... ok
test git::test_diff ... ok
test git::test_log ... FAILED
test runner::test_execute ... FAILED
test runner::test_fallback ... ok

failures:

---- git::test_log stdout ----
thread 'git::test_log' panicked at 'assertion failed: `(left == right)`
  left: `"5 commits"`,
 right: `"3 commits"`', src/git.rs:42:5
note: run with `RUST_BACKTRACE=1` for a backtrace

---- runner::test_execute stdout ----
thread 'runner::test_execute' panicked at 'Expected exit code 0, got 1', src/runner.rs:18:5

failures:
    git::test_log
    runner::test_execute

test result: FAILED. 13 passed; 2 failed; 0 ignored; 0 measured; 0 filtered out

RTK test cargo test (~20 lines)

FAILED: 2/15 tests
  ✗ git::test_log: assertion failed at src/git.rs:42
     Expected: "3 commits", Got: "5 commits"
  ✗ runner::test_execute: Expected exit code 0, got 1 at src/runner.rs:18

Token savings: 90% (180+ lines hidden)

Git Status

Standard git status (~50 lines, ~800 tokens)

On branch feature/token-optimization
Your branch is up to date with 'origin/feature/token-optimization'.

Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
	modified:   src/filter.rs
	modified:   src/git.rs
	modified:   src/tracking.rs

Untracked files:
  (use "git add <file>..." to include in what will be committed)
	src/discover/

no changes added to commit (use "git add" and/or "git commit -a")

RTK git status (1 line, ~20 tokens)

3 modified, 1 untracked ✓

Token savings: 97% (780 tokens saved)

Why Token Reduction Matters

Lower Costs

With Claude Sonnet 4.0 at

3/MTok input, saving 100K tokens per day = **

300/month savings**.

Faster Responses

Smaller context = faster LLM processing. RTK reduces latency by 20-40% on large commands.

Context Limits

Claude has a 200K token context window. RTK helps you stay under the limit on complex projects.

Better Focus

LLMs perform better with concise, relevant context. RTK removes noise and highlights failures.

Token Savings by Command Type

File Operations (50-80% savings)

Command	Strategy	Savings	Example
`rtk ls`	Tree compression	50-70%	Aggregate dirs with counts
`rtk read`	Code filtering	40-90%	Strip comments/bodies
`rtk grep`	Grouping by file	60-80%	Truncate long lines
`rtk find`	Tree hierarchy	50-70%	Compact paths

Git Operations (75-92% savings)

Command	Strategy	Savings	Example
`rtk git status`	Stats extraction	80%	“3 modified, 1 untracked”
`rtk git diff`	Stats + compact	75%	“+142/-89, 3 files”
`rtk git log`	One-line summaries	80%	“5 commits, +142/-89”
`rtk git push`	Ultra-compact	92%	“ok ✓ main”

Testing (90-99% savings)

Command	Strategy	Savings	Example
`rtk test`	Failures only	90%	Hide passing tests
`rtk vitest`	ANSI stripping	99.5%	Show only failures
`rtk pytest`	State machine	90%	Parse test outcomes
`rtk playwright`	Failure focus	94%	Group by suite

Linting (80-85% savings)

Command	Strategy	Savings	Example
`rtk lint`	Group by rule	84%	“no-unused-vars: 23”
`rtk tsc`	Group by file	83%	“errors.ts: 5 errors”
`rtk ruff`	JSON parsing	80%	Rule-based grouping
`rtk golangci-lint`	JSON parsing	85%	Rule-based grouping

Measuring Your Savings

RTK tracks all token savings automatically in SQLite (~/.local/share/rtk/history.db).

View Summary Stats

rtk gain

Example output

📊 RTK Token Savings
════════════════════════════════════════

Total commands:    133
Input tokens:      30.5K
Output tokens:     10.7K
Tokens saved:      25.3K (83.0%)

By Command:
────────────────────────────────────────
Command               Count      Saved     Avg%
rtk git status           41      17.4K    82.9%
rtk git push             54       3.4K    91.6%
rtk grep                 15       3.2K    26.5%
rtk ls                   23       1.4K    37.2%

View Daily Breakdown

rtk gain --daily

Example output

Daily Savings (last 30 days):
────────────────────────────────────────
01-23 │███████████████████                      6.4K
01-24 │██████████████████                       5.9K
01-25 │                                         18
01-26 │████████████████████████████████████████ 13.0K

Export to CSV/JSON

rtk gain --all --format csv > savings.csv
rtk gain --all --format json > savings.json

Cost Analysis Example

Scenario: Mid-size TypeScript project, 8 hours/day coding with Claude Code

Without RTK

Commands per day: ~500
Avg tokens per command: ~300
Daily token usage: 150K tokens
Monthly usage: 3.3M tokens
Monthly cost (Claude Sonnet 4.0 @ $3/MTok): **$ 9.90**

With RTK

Commands per day: ~500
Avg tokens per command: ~60 (80% reduction)
Daily token usage: 30K tokens
Monthly usage: 660K tokens
Monthly cost: $1.98
Monthly savings: $7.92 (80%)

Scales with usage: Teams with 10 developers save ~ $80/month**. Enterprise teams (100+ devs) can save **$ 800+/month.

Token Estimation Formula

RTK uses a simple heuristic for token estimation:

fn estimate_tokens(text: &str) -> usize {
    (text.len() as f64 / 4.0).ceil() as usize
}

Approximation: ~4 characters per token (GPT-style tokenization) This is a rough estimate but accurate enough for tracking purposes. Real tokenization varies by model (Claude uses a different tokenizer than GPT), but the 4:1 ratio is a good average.

Filter Effectiveness

Not all filters save the same percentage. Here’s how effectiveness varies:

High Effectiveness

90-99% savings

Test output (failures only)
Git operations (push/pull)
Package managers (install)

Medium Effectiveness

70-85% savings

Linting (grouped errors)
Git diffs (stats mode)
Directory listings

Low Effectiveness

40-60% savings

Code reading (minimal filter)
Log files (deduplication)
JSON inspection

Best Practices for Maximum Savings

Use the hook

Install rtk init -g for 100% adoption. Manual prefixing leads to ~60-70% adoption.

Prefer rtk for repetitive commands

Commands you run 10+ times per session (git status, test, lint) benefit most.

Use aggressive filter for code reading

rtk read file.rs -l aggressive strips function bodies, saving 60-90% on large files.

Let failures surface naturally

Don’t use -v flags unless debugging. Compact output is optimized for LLMs.

Review savings monthly

Run rtk gain --monthly to track trends and identify optimization opportunities.

Get Started

Core Concepts

Commands

Integration

Analytics

Advanced

Resources

30-Minute Claude Code Session Breakdown

Real Examples: Before & After

Directory Listing

Git Operations

Test Output

Git Status

Why Token Reduction Matters

Lower Costs

Faster Responses

Context Limits

Better Focus

Token Savings by Command Type

File Operations (50-80% savings)

Git Operations (75-92% savings)

Testing (90-99% savings)

Linting (80-85% savings)

Measuring Your Savings

View Summary Stats

View Daily Breakdown

Export to CSV/JSON

Cost Analysis Example

Without RTK

With RTK

Token Estimation Formula

Filter Effectiveness

High Effectiveness

Medium Effectiveness

Low Effectiveness

Best Practices for Maximum Savings

Get Started

Core Concepts

Commands

Integration

Analytics

Advanced

Resources

​30-Minute Claude Code Session Breakdown

​Real Examples: Before & After

​Directory Listing

​Git Operations

​Test Output

​Git Status

​Why Token Reduction Matters

Lower Costs

Faster Responses

Context Limits

Better Focus

​Token Savings by Command Type

​File Operations (50-80% savings)

​Git Operations (75-92% savings)

​Testing (90-99% savings)

​Linting (80-85% savings)

​Measuring Your Savings

​View Summary Stats

​View Daily Breakdown

​Export to CSV/JSON

​Cost Analysis Example

​Without RTK

​With RTK

​Token Estimation Formula

​Filter Effectiveness

High Effectiveness

Medium Effectiveness

Low Effectiveness

​Best Practices for Maximum Savings

30-Minute Claude Code Session Breakdown

Real Examples: Before & After

Directory Listing

Git Operations

Test Output

Git Status

Why Token Reduction Matters

Token Savings by Command Type

File Operations (50-80% savings)

Git Operations (75-92% savings)

Testing (90-99% savings)

Linting (80-85% savings)

Measuring Your Savings

View Summary Stats

View Daily Breakdown

Export to CSV/JSON

Cost Analysis Example

Without RTK

With RTK

Token Estimation Formula

Filter Effectiveness

Best Practices for Maximum Savings