Context Engine for AI Coding Tools · Open Source

Your AI reads 10,000 lines
to find 1 function.

Name: Indexa
Author: Prashant Singh

Indexa gives it exactly what it needs.

Index your codebase once. Return only relevant symbols, dependencies, and execution flows. 50% fewer tokens. Better answers.

Get Started View on GitHub

50-70%

Token reduction

MCP tools

CLI commands

<2s

Query time

Terminal

$ npx indexa-mcp setup

✓ Project detected: my-app (typescript/react)

✓ Indexed 8,500 chunks in 15s

✓ MCP configured for Claude Code

✓ Ready! Try: "explain the auth flow"

See It In 5 Seconds

One query. Full system understanding.

You ask:

"trace the login flow"

Indexa returns:

              VendorAuthGuard

                → VendorFlowGuard

                  → verifyPkceSession

                    → getAppSessionCookie

                      → useVcAuthStore

9 steps · 5 files · 2,500 tokens · Done.

You ask:

"what breaks if I change UserService?"

Indexa returns:

              12 references across 8 files

              Blast radius: LoginController, AuthMiddleware,

              SessionManager, TokenValidator, +4 more

Impact mapped · Before you refactor.

You ask:

"explain the theme system"

Indexa returns:

              ThemeSwitcher → useUIStore → ThemeApplier

              4 themes: default, cyberpunk, minimal, matrix

              CSS variables swap via data-theme attribute

3 symbols · 1,789 tokens · Complete picture.

Most tools help you find code. Indexa helps AI understand it.

The Problem

AI agents waste 95% of tokens reading full files

Every time an AI reads your code, it consumes thousands of tokens on imports, boilerplate, and irrelevant functions. Indexa changes that.

Without Indexa

Full File Reading

AI agent reads auth.ts (450 lines) just to understand one function.

Reads 12 files to answer a single question.

Burns through context window in minutes.

~10,000 tokens

With Indexa

Smart Code Retrieval

Returns only the relevant validateToken() function and its dependencies.

Semantic search finds exactly what the AI needs.

Context window stays efficient and focused.

~3,000 tokens

How It Works

Three steps to smarter AI coding

From install to intelligent code retrieval in under 30 seconds.

Install

One command installs Indexa globally and configures it for your AI tools.

$ npm install -g indexa-mcp

Index

Automatically parses your codebase into semantic chunks with ML embeddings.

$ npx indexa-mcp setup

Query

AI agents get exactly the code they need through 19 specialized MCP tools.

"explain the auth flow"

Features

Everything you need for intelligent code retrieval

Built for developers who want their AI agents to work smarter, not harder.

50-70% Token Reduction

Returns only relevant code chunks instead of full files. Your AI agents use fewer tokens and produce better results.

Local-First

Runs entirely on your machine. No cloud dependency, no API keys, no recurring costs. Your code never leaves your device.

ML Embeddings

Hybrid Search

Combines 35% semantic, 25% BM25, 15% name matching, and 25% path scoring for precise, relevant results.

Query Intent Classification

Automatically detects whether you want flow analysis, explanations, references, debugging, or general search.

VS Code Extension

Inline AI commands: Explain This, Fix This, Refactor, Generate Tests. Right-click menu + keyboard shortcuts. Auto-index on save.

Zero-Config Setup

npx indexa-mcp setup — indexes, configures MCP, creates CLAUDE.md, verifies. Claude Code uses Indexa automatically.

Code Quality Built In

Dead code detection, circular dependencies, unused exports, duplicate code finder, security scan — all in one tool.

Battle-Tested Reliability

Binary file protection, corrupt index recovery, embedding fallback, buffer overflow guards. 23 integration tests. Built for daily use.

Token Savings

Real numbers, real savings

Measured across real-world codebases. Indexa consistently delivers 50-70% token reduction.

Query Type	Without Indexa	With Indexa	Savings
Explain a function	~8,500 tokens	~2,800 tokens	67%
Trace auth flow	~15,000 tokens	~5,200 tokens	65%
Find references	~12,000 tokens	~4,100 tokens	66%
Debug an issue	~10,000 tokens	~3,500 tokens	65%
Search for patterns	~6,000 tokens	~2,100 tokens	65%
Average	~10,300 tokens	~3,540 tokens	66%

MCP Tools

19 specialized tools for every query

Search, trace, analyze, and review code. Dead code detection, circular deps, duplicate finder, security scans, PR review, and more.

context_bundle

Smart context packaging that bundles only the relevant code chunks for a query.

Hybrid semantic + keyword search across your entire codebase with ranked results.

explain

Retrieves code with surrounding context optimized for explanation queries.

flow

Traces execution flow across files, following function calls and data paths.

symbol

Looks up specific symbols — functions, classes, interfaces, types — by name.

file

Returns the full content of a specific file when complete context is needed.

references

Finds all references and usages of a symbol across your project files.

index

Re-indexes your codebase to pick up new changes and update embeddings.

stats

Shows indexing statistics — chunk counts, file coverage, embedding status.

dead_code

Find unreferenced functions, methods, and classes across the entire codebase.

blast_radius

What breaks if you change a symbol? Direct refs, transitive impact, affected files.

circular_deps

Detect circular dependencies between files. Break import cycles before they break you.

duplicates

Find near-duplicate code via embedding similarity. Spot copy-paste patterns instantly.

code_grep

Regex pattern search across all indexed source files. Find console.log, TODO, hardcoded URLs.

security_scan

Deep security analysis grouped by OWASP domains — auth, XSS, secrets, crypto.

review_pr

Context-aware PR review — changed files, blast radius, connections, review checklist.

impact_chain

Full transitive impact analysis — trace every symbol affected across all depths.

Quick Start

Up and running in 30 seconds

Three commands. That is all it takes.

              bash
              
# Install globally
$ npm install -g indexa-mcp

# Setup in your project (detects, indexes, configures MCP)
$ npx indexa-mcp setup

# Or run directly without installing
$ npx indexa-mcp setup

Claude Code MCP config (~/.mcp.json)

{
  "mcpServers": {
    "indexa": {
      "command": "npx",
      "args": ["-y", "indexa-mcp", "stdio"]
    }
  }
}
            

Architecture

How Indexa processes your code

A lightweight pipeline that transforms raw source files into semantically searchable chunks.

Input

Source Files

Parse

AST Chunker

Embed
Transformers.js

Store

Local Index

Serve
MCP Server

Smart Parsing

AST-aware chunking preserves function boundaries and semantic meaning

384-dim Embeddings

GTE-small model via Transformers.js for semantic understanding

Hybrid Ranking

Semantic + BM25 + name + path scoring for best results

100% Local

Everything runs on your machine. Zero network calls required

Chunks indexed

Token reduction verified

MCP tools available

Average query time

Comparison

Why not just use RAG / Copilot / Grep?

Different tools. Different capabilities. See where Indexa wins.

Capability	Generic RAG	Copilot	Grep / Glob	Indexa
Returns	Raw text chunks	Full file reads	Line matches	Symbols + deps + connections
Understands structure	✗	✗	✗	✓ AST-parsed
Traces execution	✗	✗	✗	✓ Call chains across files
Blast radius	✗	✗	✗	✓ What breaks if you change X
Dead code detection	✗	✗	✗	✓ Unreferenced symbols
Circular dependencies	✗	✗	✗	✓ Import cycle detection
Duplicate code finder	✗	✗	✗	✓ Embedding similarity
Token efficiency	Poor	Poor	Medium	50-70% reduction
Runs locally	Sometimes	✗	✓	✓ Always
Free	Sometimes	✗	✓	Free forever

Creator

Built by Prashant Singh

Module Lead & Frontend–AI Specialist with 6+ years building scalable systems.

Prashant Singh

Module Lead | Frontend – AI & Migration Specialist

Experienced Senior Frontend Engineer skilled in full-stack development, API integration, and Agile methodologies. Proficient in Angular, React, Next.js, TypeScript, and modern DevOps. Expert in building scalable, high-performance web applications with focus on AI-powered tooling.

6+Years Exp.

50+Features

35+Projects

847Contributions

AngularJS → React migrations at scale

AI tools that cut dev effort by 40%

Performance wins: LCP 4.2s → 1.8s

Systems serving 100K+ users

GitHub LinkedIn Portfolio

Like Indexa? Give it a star

Help other developers discover Indexa. A star on GitHub goes a long way.

Star on GitHub

Start saving tokens today

One command. Zero configuration. Free forever.

Get Started View on npm Star on GitHub

Your AI reads 10,000 linesto find 1 function.