LCI - Lightning Code Index

Lightning-fast code indexing and search for AI assistants.

Features

• Sub-millisecond search: Trigram-based indexing with <5ms search guarantee
• Multi-language support: Go, TypeScript, JavaScript, Python, Rust, C#, PHP, and more
• MCP integration: Model Context Protocol server for AI assistant integration
• Semantic search: Natural language queries with intelligent matching
• Call graph analysis: Track function calls, references, and dependencies
• Semantic annotations: @lci: vocabulary for marking up code with metadata

Installation

npm (recommended)

npm install -g @standardbeagle/lci

pip

pip install lightning-code-index

Homebrew (coming soon)

brew install standardbeagle/tap/lci

From Source

go install github.com/standardbeagle/lci/cmd/lci@latest

From Releases

Download pre-built binaries from GitHub Releases.

Quick Start

CLI Usage

# Index and search in current directory
lci search "handleRequest"

# Find symbol definitions
lci def UserService

# Find all references to a symbol
lci refs CreateUser

# Display function call hierarchy
lci tree main

# Fast grep-style search
lci grep "TODO|FIXME"

# List files that would be indexed
lci list

MCP Server

Start the MCP server for AI assistant integration:

lci mcp

Claude Code Integration

Add to your .mcp.json:

{
  "lci": {
    "command": "lci",
    "args": ["mcp"],
    "env": {}
  }
}

Configuration

Create .lci.kdl in your project root:

project {
    name "my-project"
    root "."
}

index {
    include "**/*.go" "**/*.ts" "**/*.py"
    exclude "**/node_modules/**" "**/vendor/**"
}

search {
    max-results 100
    context-lines 3
}

MCP Tools

When running as an MCP server, LCI exposes these tools:

| Tool | Description | |------|-------------| | search | Semantic code search with fuzzy matching | | get_context | Get detailed context for a code symbol | | find_files | Fast file path search with glob patterns | | code_insight | Codebase intelligence and analysis | | context | Save/load code context manifests | | semantic_annotations | Query @lci: semantic labels | | side_effects | Analyze function purity and side effects |

Semantic Annotations

Mark up your code with @lci: annotations for enhanced AI understanding:

// @lci:risk[high] @lci:public-api
// @lci:requires[env:DATABASE_URL]
func HandleUserLogin(w http.ResponseWriter, r *http.Request) {
    // ...
}

// @lci:purpose[Validate user credentials against database]
// @lci:must[Return error for invalid credentials]
func ValidateCredentials(username, password string) error {
    // ...
}

Annotation Categories

• Risk & Safety: @lci:risk[low|medium|high|critical], @lci:safe-zone, @lci:stability
• Dependencies: @lci:requires[env:VAR], @lci:requires[db:table], @lci:requires[service:name]
• Conventions: @lci:convention[pattern], @lci:idiom[name], @lci:template[name]
• Contracts: @lci:must[behavior], @lci:must-not[behavior], @lci:invariant[condition]
• Purpose: @lci:purpose[description], @lci:domain[area], @lci:owner[team]

Architecture

lci/
├── cmd/lci/          # CLI entry point
├── internal/
│   ├── core/         # Trigram index, symbol store, reference tracker
│   ├── parser/       # Tree-sitter based multi-language parsing
│   ├── search/       # Search engine and scoring
│   ├── indexing/     # Master index and pipeline
│   ├── mcp/          # MCP server and tools
│   └── analysis/     # Code analysis and metrics
└── pkg/              # Public API

Performance

LCI is designed for speed:

• Indexing: <5s for typical projects (<10k files)
• Search: <5ms for most queries
• Memory: <100MB for typical web projects
• Startup: Near-instant with persistent index

Development

# Run tests
go test ./...

# Build
go build ./cmd/lci

# Run with race detector
go test -race ./...

License

MIT License - see LICENSE for details.

Contributing

Contributions are welcome! Please read the contributing guidelines before submitting PRs.