AiDex

Stop wasting 80% of your AI's context window on code searches.

AiDex is an MCP server that gives AI coding assistants instant access to your entire codebase through a persistent, pre-built index. Works with any MCP-compatible AI assistant: Claude Code, Claude Desktop, Cursor, Windsurf, Gemini CLI, VS Code Copilot, and more.

What's Inside — 30 Tools in One Server

11 languages — C#, TypeScript, JavaScript, Rust, Python, C, C++, Java, Go, PHP, Ruby

# Find where "PlayerHealth" is defined — 1 call, ~50 tokens
aidex_query({ term: "PlayerHealth" })
→ Engine.cs:45, Player.cs:23, UI.cs:156

# All methods in a file — without reading the whole file
aidex_signature({ file: "src/Engine.cs" })
→ class GameEngine { Update(), Render(), LoadScene(), ... }

# What changed in the last 2 hours?
aidex_query({ term: "render", modified_since: "2h" })

# Search across ALL your projects at once
aidex_global_query({ term: "TransparentWindow", mode: "contains" })
→ Found in: LibWebAppGpu (3 hits), DebugViewer (1 hit)

# Leave a note for your next session
aidex_note({ path: ".", note: "Test the parser fix after restart" })

# Create a task while working
aidex_task({ path: ".", action: "create", title: "Fix edge case in parser", priority: 1, tags: "bug" })

Table of Contents

• What's Inside
• The Problem
• The Solution
• Why Not Just Grep?
• How It Works
• Features
• Supported Languages
• Quick Start
• Available Tools
• Time-based Filtering
• Project Structure
• Session Notes
• Task Backlog
• Global Search
• AI Guidelines
• Log Hub
• Screenshots — LLM-Optimized
• Interactive Viewer
• CLI Usage
• Performance
• Technology
• Contributing
• License

The Problem

Every time your AI assistant searches for code, it:

• Greps through thousands of files → hundreds of results flood the context
• Reads file after file to understand the structure → more context consumed
• Forgets everything when the session ends → repeat from scratch

A single "Where is X defined?" question can eat 2,000+ tokens. Do that 10 times and you've burned half your context on navigation alone.

The Solution

Index once, query forever:

# Before: grep flooding your context
AI: grep "PlayerHealth" → 200 hits in 40 files
AI: read File1.cs, File2.cs, File3.cs...
→ 2000+ tokens consumed, 5+ tool calls

# After: precise results, minimal context
AI: aidex_query({ term: "PlayerHealth" })
→ Engine.cs:45, Player.cs:23, UI.cs:156
→ ~50 tokens, 1 tool call

Result: 50-80% less context used for code navigation.

Why Not Just Grep?

The real cost of grep: Every grep result includes surrounding context. Search for User in a large project and you'll get hundreds of hits - comments, strings, partial matches. Your AI reads through all of them, burning context tokens on noise.

AiDex indexes identifiers: It uses Tree-sitter to actually parse your code. When you search for User, you get the class definition, the method parameters, the variable declarations - not every comment that mentions "user".

How It Works

1. Index your project once (~1 second per 1000 files)

   code

   aidex_init({ path: "/path/to/project" })
   

2. AI searches the index instead of grepping

   code

   aidex_query({ term: "Calculate", mode: "starts_with" })
   → All functions starting with "Calculate" + exact line numbers
   
   aidex_query({ term: "Player", modified_since: "2h" })
   → Only matches changed in the last 2 hours
   

3. Get file overviews without reading entire files

   code

   aidex_signature({ file: "src/Engine.cs" })
   → All classes, methods, and their signatures
   

The index lives in .aidex/index.db (SQLite) - fast, portable, no external dependencies.

Features

• Tree-sitter Parsing: Real code parsing, not regex — indexes identifiers, ignores keywords and noise
• ~50 Tokens per Search: vs 2000+ with grep — your AI keeps its context for actual work
• Persistent Index: Survives between sessions — no re-scanning, no re-reading
• Incremental Updates: Re-index single files after changes, not the whole project
• Time-based Filtering: Find what changed in the last hour, day, or week
• Auto-Cleanup: Excluded files (e.g., build outputs) are automatically removed from index
• Zero Dependencies: SQLite with WAL mode — single file, fast, portable

Supported Languages

Quick Start

1. Install

npm install -g aidex-mcp

That's it. Setup runs automatically after install — it detects your installed AI clients (Claude Code, Claude Desktop, Cursor, Windsurf, Gemini CLI, VS Code Copilot) and registers AiDex as an MCP server. It also adds usage instructions to your AI's config (~/.claude/CLAUDE.md, ~/.gemini/GEMINI.md).

To re-run setup manually: aidex setup | To unregister: aidex unsetup | To skip auto-setup: AIDEX_NO_SETUP=1 npm install -g aidex-mcp

2. Or register manually with your AI assistant

For Claude Code (~/.claude/settings.json or ~/.claude.json):

{
  "mcpServers": {
    "aidex": {
      "type": "stdio",
      "command": "aidex",
      "env": {}
    }
  }
}

For Claude Desktop (%APPDATA%/Claude/claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "aidex": {
      "command": "aidex"
    }
  }
}

Note: Both aidex and aidex-mcp work as command names.

Important: The server name in your config determines the MCP tool prefix. Use "aidex" as shown above — this gives you tool names like aidex_query, aidex_signature, etc. Using a different name (e.g., "codegraph") would change the prefix accordingly.

For Gemini CLI (~/.gemini/settings.json):

{
  "mcpServers": {
    "aidex": {
      "command": "aidex"
    }
  }
}

For VS Code Copilot (run MCP: Open User Configuration in Command Palette):

{
  "servers": {
    "aidex": {
      "type": "stdio",
      "command": "aidex"
    }
  }
}

For other MCP clients: See your client's documentation for MCP server configuration.

3. Make your AI actually use it

Add to your AI's instructions (e.g., ~/.claude/CLAUDE.md for Claude Code, or the equivalent for your AI client). This tells the AI when and how to use AiDex instead of grepping:

## AiDex - Persistent Code Index (MCP Server)

AiDex provides fast, precise code search through a pre-built index.
**Always prefer AiDex over Grep/Glob for code searches.**

### REQUIRED: Before using Grep/Glob/Read for code searches

Do I want to search code? ├── .aidex/ exists → STOP! Use AiDex instead ├── .aidex/ missing → run aidex_init (don't ask), THEN use AiDex └── Config/Logs/Text → Grep/Read is fine

**NEVER do this when .aidex/ exists:**
- ❌ `Grep pattern="functionName"` → ✅ `aidex_query term="functionName"`
- ❌ `Grep pattern="class.*Name"` → ✅ `aidex_query term="Name" mode="contains"`
- ❌ `Read file.cs` to see methods → ✅ `aidex_signature file="file.cs"`
- ❌ `Glob pattern="**/*.cs"` + Read → ✅ `aidex_signatures pattern="**/*.cs"`

### Session-Start Rule (REQUIRED — every session, no exceptions)

1. Call `aidex_session({ path: "<project>" })` — detects external changes, auto-reindexes
2. If `.aidex/` does NOT exist → run `aidex_init` automatically (don't ask)
3. If a session note exists → **show it to the user** before continuing
4. **Before ending a session:** always leave a note about what to do next

### Question → Right Tool

| Question | Tool |
|----------|------|
| "Where is X defined?" | `aidex_query term="X"` |
| "Find anything containing X" | `aidex_query term="X" mode="contains"` |
| "All functions starting with X" | `aidex_query term="X" mode="starts_with"` |
| "What methods does file Y have?" | `aidex_signature file="Y"` |
| "Explore all files in src/" | `aidex_signatures pattern="src/**"` |
| "Project overview" | `aidex_summary` + `aidex_tree` |
| "What changed recently?" | `aidex_query term="X" modified_since="2h"` |
| "What files changed today?" | `aidex_files path="." modified_since="8h"` |
| "Have I ever written X?" | `aidex_global_query term="X" mode="contains"` |
| "Which project has class Y?" | `aidex_global_signatures term="Y" kind="class"` |
| "All indexed projects?" | `aidex_global_status` |

### Search Modes

- **`exact`** (default): Finds only the exact identifier — `log` won't match `catalog`
- **`contains`**: Finds identifiers containing the term — `render` matches `preRenderSetup`
- **`starts_with`**: Finds identifiers starting with the term — `Update` matches `UpdatePlayer`, `UpdateUI`

### All Tools (30)

| Category | Tools | Purpose |
|----------|-------|---------|
| Search & Index | `aidex_init`, `aidex_query`, `aidex_update`, `aidex_remove`, `aidex_status` | Index project, search identifiers (exact/contains/starts_with), time filter |
| Signatures | `aidex_signature`, `aidex_signatures` | Get classes + methods without reading files |
| Overview | `aidex_summary`, `aidex_tree`, `aidex_describe`, `aidex_files` | Entry points, file tree, file listing by type |
| Cross-Project | `aidex_link`, `aidex_unlink`, `aidex_links`, `aidex_scan` | Link dependencies, discover projects |
| Global Search | `aidex_global_init`, `aidex_global_query`, `aidex_global_signatures`, `aidex_global_status`, `aidex_global_refresh` | Search across ALL projects |
| Guidelines | `aidex_global_guideline` | Persistent AI instructions & conventions (key-value, global) |
| Sessions | `aidex_session`, `aidex_note` | Track sessions, leave notes (with searchable history) |
| Tasks | `aidex_task`, `aidex_tasks` | Built-in backlog with priorities, tags, summaries, auto-logged history |
| Log Hub | `aidex_log` | Universal log receiver — any program sends logs via HTTP, AI queries them, live in Viewer |
| Screenshots | `aidex_screenshot`, `aidex_windows` | Screen capture with LLM optimization (scale + color reduction, no index needed) |
| Viewer | `aidex_viewer` | Interactive browser UI with file tree, signatures, tasks, and live logs |

**11 languages:** C#, TypeScript, JavaScript, Rust, Python, C, C++, Java, Go, PHP, Ruby

### Session Notes

Leave notes for the next session — they persist in the database:

aidex_note({ path: ".", note: "Test the fix after restart" }) # Write aidex_note({ path: ".", note: "Also check edge cases", append: true }) # Append aidex_note({ path: "." }) # Read aidex_note({ path: ".", search: "parser" }) # Search history aidex_note({ path: ".", clear: true }) # Clear

- **Before ending a session:** automatically leave a note about next steps
- **User says "remember for next session: ..."** → write it immediately

### Task Backlog

Track TODOs, bugs, and features right next to your code index:

aidex_task({ path: ".", action: "create", title: "Fix bug", priority: 1, tags: "bug" }) aidex_task({ path: ".", action: "update", id: 1, status: "done" }) aidex_task({ path: ".", action: "log", id: 1, note: "Root cause found" }) aidex_tasks({ path: ".", status: "active" })

Priority: 1=high, 2=medium, 3=low | Status: `backlog → active → done | cancelled`

### Global Search (across all projects)

aidex_global_init({ path: "/path/to/all/repos" }) # Scan & register aidex_global_init({ path: "...", index_unindexed: true }) # + auto-index small projects aidex_global_query({ term: "TransparentWindow", mode: "contains" }) # Search everywhere aidex_global_signatures({ term: "Render", kind: "method" }) # Find methods everywhere aidex_global_status({ sort: "recent" }) # List all projects

### Screenshots

aidex_screenshot() # Full screen aidex_screenshot({ mode: "active_window" }) # Active window aidex_screenshot({ mode: "window", window_title: "VS Code" }) # Specific window aidex_screenshot({ scale: 0.5, colors: 2 }) # B&W, half size (ideal for LLM) aidex_screenshot({ colors: 16 }) # 16 colors (UI readable) aidex_windows({ filter: "chrome" }) # Find window titles

No index needed. Returns file path → use `Read` to view immediately.

**LLM optimization strategy:** Always start with aggressive settings, then retry if unreadable:
1. First try: `scale: 0.5, colors: 2` (B&W, half size — smallest possible)
2. If unreadable: retry with `colors: 16` (adds shading for UI elements)
3. If still unclear: `scale: 0.75` or omit `colors` for full quality
4. **Remember** what works for each window/app during the session — don't retry every time.

4. Index your project

Ask your AI: "Index this project with AiDex"

Or manually in the AI chat:

aidex_init({ path: "/path/to/your/project" })

Available Tools

Time-based Filtering

Track what changed recently with modified_since and modified_before:

aidex_query({ term: "render", modified_since: "2h" })   # Last 2 hours
aidex_query({ term: "User", modified_since: "1d" })     # Last day
aidex_query({ term: "API", modified_since: "1w" })      # Last week

Supported formats:

• Relative: 30m (minutes), 2h (hours), 1d (days), 1w (weeks)
• ISO date: 2026-01-27 or 2026-01-27T14:30:00

Perfect for questions like "What did I change in the last hour?"

Project Structure

AiDex indexes ALL files in your project (not just code), letting you query the structure:

aidex_files({ path: ".", type: "config" })  # All config files
aidex_files({ path: ".", type: "test" })    # All test files
aidex_files({ path: ".", pattern: "**/*.md" })  # All markdown files
aidex_files({ path: ".", modified_since: "30m" })  # Changed this session

File types: code, config, doc, asset, test, other, dir

Use modified_since to find files changed in this session - perfect for "What did I edit?"

Session Notes

Leave reminders for the next session - no more losing context between chats:

aidex_note({ path: ".", note: "Test the glob fix after restart" })  # Write
aidex_note({ path: ".", note: "Also check edge cases", append: true })  # Append
aidex_note({ path: "." })                                              # Read
aidex_note({ path: ".", clear: true })                                 # Clear

Note History (v1.10): Old notes are automatically archived when overwritten or cleared. Browse and search past notes:

aidex_note({ path: ".", history: true })                    # Browse archived notes (shows summaries)
aidex_note({ path: ".", search: "parser" })                 # Search note history (searches summaries too)
aidex_note({ path: ".", history: true, limit: 5 })          # Last 5 archived notes

Note Summaries (v1.15): Provide a summary when writing/clearing a note — the archived note gets this one-sentence description. History then shows summaries instead of truncated text:

aidex_note({ path: ".", note: "New focus", summary: "Previous session: finished parser refactoring" })

Use cases:

• Before ending a session: "Remember to test X next time"
• AI auto-reminder: Save what to verify after a restart
• Handover notes: Context for the next session without editing config files
• Search past sessions: "What did we do about the parser?"

Notes are stored in the SQLite database (.aidex/index.db) and persist indefinitely.

Task Backlog

Keep your project tasks right next to your code index - no Jira, no Trello, no context switching:

aidex_task({ path: ".", action: "create", title: "Fix parser bug", priority: 1, tags: "bug", summary: "Parser crashes on nested generics in C#" })
aidex_task({ path: ".", action: "update", id: 1, status: "done" })
aidex_task({ path: ".", action: "log", id: 1, note: "Root cause: unbounded buffer" })
aidex_tasks({ path: ".", status: "active" })

Features:

• Summaries: One-sentence table-of-contents per task — scan the backlog without reading full details
• Priorities: 🔴 high, 🟡 medium, ⚪ low
• Statuses: backlog → active → done | cancelled
• Tags: Categorize tasks (bug, feature, docs, etc.)
• History log: Every status change is auto-logged, plus manual notes
• Viewer integration: Tasks tab in the browser viewer with live updates
• Persistent: Tasks survive between sessions, stored in .aidex/index.db

Your AI assistant can create tasks while working ("found a bug in the parser, add it to the backlog"), track progress, and pick up where you left off next session.

Global Search

Search across ALL your indexed projects at once. Perfect for "Have I ever written a transparent window?" or "Where did I use that algorithm?"

Setup

aidex_global_init({ path: "Q:/develop" })                              # Scan & register
aidex_global_init({ path: "Q:/develop", exclude: ["llama.cpp"] })      # Skip external repos
aidex_global_init({ path: "Q:/develop", index_unindexed: true })       # Auto-index all found projects
aidex_global_init({ path: "Q:/develop", index_unindexed: true, show_progress: true })  # With browser progress UI

This scans your project directory, registers all AiDex-indexed projects in a global database (~/.aidex/global.db), and reports any unindexed projects it finds by detecting project markers (.csproj, package.json, Cargo.toml, etc.).

With index_unindexed: true, it also auto-indexes all discovered projects with ≤500 code files. Larger projects are listed separately for user decision. Add show_progress: true to open a live progress UI in your browser (http://localhost:3334).

Search

aidex_global_query({ term: "TransparentWindow" })                      # Exact match
aidex_global_query({ term: "transparent", mode: "contains" })          # Fuzzy search
aidex_global_signatures({ term: "Render", kind: "method" })            # Find methods
aidex_global_signatures({ term: "Player", kind: "class" })             # Find classes

How it works

• Uses SQLite ATTACH DATABASE to query project databases directly — no data copying
• Results are cached in memory (5-minute TTL) for fast repeated queries
• Projects are batched (8 at a time) to respect SQLite's attachment limit
• Each project keeps its own .aidex/index.db as the single source of truth
• Auto-deduplication: Parent projects that contain sub-projects are automatically skipped (e.g., MyApp/ is removed when MyApp/Frontend/ and MyApp/Backend/ exist as separate indexed projects)

Management

aidex_global_status()                                                  # List all projects
aidex_global_status({ sort: "recent" })                                # Most recently indexed first
aidex_global_refresh()                                                 # Update stats, remove stale

AI Guidelines

Store persistent coding conventions, review checklists, and AI instructions in a single place — shared across all projects.

aidex_global_guideline({ action: "set", key: "review", value: "Always check: error handling, null safety, no hardcoded strings" })
aidex_global_guideline({ action: "set", key: "style", value: "Use PascalCase for classes, camelCase for methods, 4-space indent" })
aidex_global_guideline({ action: "get", key: "review" })               # Retrieve a guideline
aidex_global_guideline({ action: "list" })                             # Show all guidelines
aidex_global_guideline({ action: "list", filter: "code" })             # Filter by name
aidex_global_guideline({ action: "delete", key: "old-rule" })          # Remove a guideline

Use cases:

• Code review checklist: Tell your AI exactly what to look for every time
• Coding conventions: Store team style rules once, reference them in any project
• Release checklist: Step-by-step process for shipping
• Project-agnostic instructions: No more pasting the same context into every session

Guidelines are stored in ~/.aidex/global.db — available across all your projects without aidex_init. Ask your AI: "Load the review guideline and apply it to this file."

Log Hub — Universal Logging

Turn any program into a log source for your AI assistant. Your app sends logs via HTTP POST, the AI queries them via MCP, and you see them live in the Viewer — zero dependencies, zero setup in your code.

How it works

Your Program ──HTTP POST──→ AiDex Log Hub (port 3335) ──→ Ring Buffer
                                        │                      │
                                        │ WebSocket             │ MCP query
                                        ↓                      ↓
                                   Viewer (Logs tab)      AI Assistant
                                   (you see live)       (queries & analyzes)

Quick start

1. AI starts the Log Hub: aidex_log({ action: "init" })
2. AI opens the Viewer: aidex_viewer({ path: "." }) — Logs tab shows live stream
3. Add one line to your program:

// C#
await new HttpClient().PostAsJsonAsync("http://localhost:3335/log",
    new { level = "info", source = "MyApp", message = "Player spawned", data = new { x = 10, y = 20 } });

# Python
requests.post("http://localhost:3335/log", json={"level": "info", "source": "MyApp", "message": "Done"})

// JavaScript
fetch("http://localhost:3335/log", {
    method: "POST", headers: {"Content-Type": "application/json"},
    body: JSON.stringify({level: "info", source: "MyApp", message: "Started"})
});

# PowerShell
Invoke-RestMethod -Uri http://localhost:3335/log -Method POST -ContentType "application/json" -Body '{"level":"info","source":"Script","message":"Done"}'

HTTP API

Fields: level (debug/info/warn/error), source (app name), message (text, required), data (optional JSON), timestamp (optional, ms)

Features

• Ring Buffer: Fixed-size in-memory FIFO (default 10,000 entries) — oldest entries overwritten
• Zero-cost: No server, no buffer, no resources until init is called
• Persistence: Optional SQLite storage with 7-day auto-cleanup (persist: true)
• Consume pattern: query with consume: true removes returned entries — ideal for polling
• Viewer integration: Logs tab with WebSocket live-stream, level/source/text filters, auto-scroll
• Fire & forget: Just POST and go — if the server isn't running, the POST silently fails

Screenshots — LLM-Optimized

Take screenshots and reduce them up to 95% for LLM context. A typical screenshot goes from ~100 KB to ~5 KB — that's thousands of tokens saved per image.

Why this matters

Most screenshots in AI context are for reading text — error messages, logs, UI labels. You don't need 16 million colors for that.

Usage

aidex_screenshot()                                             # Full screen (full quality)
aidex_screenshot({ mode: "active_window" })                    # Active window
aidex_screenshot({ mode: "window", window_title: "VS Code" }) # Specific window
aidex_screenshot({ scale: 0.5, colors: 2 })                   # B&W, half size (best for text)
aidex_screenshot({ scale: 0.5, colors: 16 })                  # 16 colors (UI readable)
aidex_screenshot({ colors: 256 })                              # 256 colors (good quality)
aidex_screenshot({ mode: "region" })                           # Interactive selection
aidex_screenshot({ mode: "rect", x: 100, y: 200, width: 800, height: 600 })  # Coordinates
aidex_windows({ filter: "chrome" })                            # Find window titles

Optimization parameters

Recommended strategy for AI assistants

The tool description tells LLMs to optimize automatically:

1. Start aggressive: scale: 0.5, colors: 2 (smallest possible)
2. If unreadable: retry with colors: 16 (adds shading for UI elements)
3. If still unclear: try scale: 0.75 or full color
4. Remember: cache what works per window/app for the rest of the session

This way the AI learns the right settings per app without wasting tokens on oversized images.

Features

• 5 capture modes: Fullscreen, active window, specific window (by title), interactive region selection, coordinate-based rectangle
• Cross-platform: Windows (PowerShell + System.Drawing), macOS (sips + ImageMagick), Linux (ImageMagick)
• Multi-monitor: Select which monitor to capture
• Delay: Wait N seconds before capturing (e.g., to open a menu first)
• Size reporting: Shows original → optimized size and percentage saved
• Auto-path: Default saves to temp directory with fixed filename
• No index required: Works standalone, no .aidex/ needed

Interactive Viewer

Explore your indexed project visually in the browser:

aidex_viewer({ path: "." })

Opens http://localhost:3333 with:

• Interactive file tree - Click to expand directories
• File signatures - Click any file to see its types and methods
• Live reload - Changes detected automatically while you code
• Git status icons - See which files are modified, staged, or untracked
• Logs tab - Live log stream from Log Hub with filters (level, source, text search)
• Tasks tab - View and manage your task backlog

Close with aidex_viewer({ path: ".", action: "close" })

CLI Usage

aidex scan Q:/develop       # Find all indexed projects
aidex init ./myproject      # Index a project from command line

aidex-mcp works as an alias for aidex.

Performance

Technology

• Parser: Tree-sitter - Real parsing, not regex
• Database: SQLite with WAL mode - Fast, single file, zero config
• Protocol: MCP - Works with any compatible AI

Project Structure

.aidex/                  ← Created in YOUR project
├── index.db             ← SQLite database
└── summary.md           ← Optional documentation

AiDex/                   ← This repository
├── src/
│   ├── commands/        ← Tool implementations
│   ├── db/              ← SQLite wrapper
│   ├── parser/          ← Tree-sitter integration
│   └── server/          ← MCP protocol handler
└── build/               ← Compiled output

Community

GitHub Discussions — Ask questions, share your setup, suggest ideas.

Contributing

See CONTRIBUTING.md for full details. Quick summary:

• Bug? → Open an issue
• Idea? → Start a Discussion
• New language? → Add a keyword file in src/parser/languages/ and open a PR

License

MIT License - see LICENSE

Authors

Uwe Chalas & Claude