rn-token-optimizer

Cut AI token waste in React Native and .NET projects.
Build a local code intelligence graph, compress prompts and terminal output, and wire it into Cursor, Kiro, or Claude Desktop via MCP — no cloud upload, no extra API key required.

Changelog · Live site · Full guide · Setup & MCP

---

What this project does

AI coding agents burn tokens in three predictable ways: verbose prompts, noisy terminal logs, and reading files one-by-one to answer structural questions. rn-token-optimizer addresses all three locally.

Capability	What you get
Code intelligence graph	Parses TypeScript/JavaScript (React Native) and C# (.NET via Roslyn) into a persistent SQLite graph of functions, screens, controllers, services, and call relationships
Semantic search	Find code by concept — "auth flow", "payment error" — without knowing exact symbol names (TF-IDF)
Visual explorer	graph ui opens http://localhost:7842 — interactive force-directed dashboard with search, filters, and source snippets
Live watcher	graph watch re-indexes on save so the graph stays current while you code
Prompt optimization	Compresses natural-language prompts into a compact Military-English DSL — typically 70–90% fewer tokens
Terminal compression	Pipes Metro, Jest, or build output through compression and returns a short DSL answer — typically 95–99% fewer tokens
MCP server	19 tools for your IDE agent: index, search, trace calls, detect diff impact, optimize prompts, and more

Everything runs on your machine. Graph data lives in .rn-token-optimizer/graph.db per project. In passthrough mode (default in Cursor/Kiro), compression uses your existing IDE subscription — no Anthropic key needed.

---

The problem (in one example)

Verbose prompt — 44 tokens:

I need you to please help me fix the issue in my React Native app where users
are getting an error when trying to log in with their Google account on Android
devices running version 12 or higher

Compressed DSL — 6 tokens:

D=fix A Google login fail AND v12+

File-by-file exploration — an agent reads ~15 files (~18,000 tokens) to answer “who calls handleGoogleLogin?” A single graph query answers the same question in ~300 tokens.

---

Quick start

React Native (one command)

cd /path/to/your-react-native-app
npx rn-token-optimizer setup

setup detects your IDE, writes MCP config, indexes the project, and prints next steps.

.NET / C# project

# Prerequisites: Node.js >= 18 and .NET 8 SDK
npm install -g rn-token-optimizer
cd /path/to/YourDotNetApp
rn-token-optimizer graph index

The tool auto-detects .csproj / .sln files and uses the Roslyn-backed indexer. First run builds the analyzer once (~30 s); later runs reuse the cached DLL.

Visual graph explorer

npm install -g rn-token-optimizer
cd /path/to/your-project

rn-token-optimizer graph index    # build SQLite graph
rn-token-optimizer graph ui       # → http://localhost:7842

Install globally (both stacks)

npm install -g rn-token-optimizer
rn-token-optimizer setup

One-line install script

curl -fsSL https://raw.githubusercontent.com/Pradeep241094/rn-token-optimizer/main/install.sh | bash

---

Use in your IDE

After setup, ask your agent naturally:

"Index my project"
"Show my app architecture"
"Who calls handleGoogleLogin?"
"Find all auth-related code"
"What does my current git diff break?"

For prompt optimization in Cursor/Kiro, start a message with the keyword:

rn-token-optimizer Fix the Google login issue on Android 12 devices

IDE	Config written
Cursor	.cursor/mcp.json + .cursor/rules/rn-token-optimizer.mdc
Kiro	.kiro/settings/mcp.json + steering file
Claude Desktop	claude_desktop_config.json (direct API key recommended)

See SETUP.md for MCP wiring, DSL reference, and troubleshooting.

---

CLI essentials

# Prompt & terminal
rn-token-optimizer optimize "your verbose prompt"
npx jest 2>&1 | rn-token-optimizer "Which tests failed?"

# Graph
rn-token-optimizer graph index
rn-token-optimizer graph architecture
rn-token-optimizer graph search "Login" --label Screen
rn-token-optimizer graph trace handleGoogleLogin --direction inbound
rn-token-optimizer graph semantic "authentication flow"
rn-token-optimizer graph dead-code
rn-token-optimizer graph changes
rn-token-optimizer graph ui
rn-token-optimizer graph watch

# Project context & DSL
rn-token-optimizer index
rn-token-optimizer dsl show

Full command list and MCP tool reference: GUIDE.md.

---

Sample projects

Sample	Path	Try it
React Native demo	demo-rn-app/	rn-token-optimizer graph index --dir ./demo-rn-app then graph ui
.NET Todo API	samples/dotnet-todo-app/	dotnet run then rn-token-optimizer graph index — see GRAPH_WALKTHROUGH.md

---

Programmatic API

import { optimizePrompt, aiTokenOptimizer, countTokensSync } from 'rn-token-optimizer';

const result = await optimizePrompt("Fix Google login on Android 12", {
  apiKey: process.env.ANTHROPIC_API_KEY, // optional in passthrough mode
});
console.log(result.optimizedPrompt); // D=fix A Google login fail AND v12+

Types and graph exports are documented in GUIDE.md.

---

Documentation map

Doc	Contents
GUIDE.md	Complete feature walkthrough, CLI reference, MCP tools, demos
SETUP.md	Installation, IDE integration, DSL system, configuration
CHANGELOG.md	Version history
samples/dotnet-todo-app/GRAPH_WALKTHROUGH.md	Step-by-step .NET graph demo

---

License

MIT — Copyright © 2026 Pradeep Margasahayam Prakash