Getting Started

Go from zero to your first semantic query in under 5 minutes.

Prerequisites

  • macOS, Linux, or Windows (via PowerShell)
  • A codebase you want to index

Installation

Run the install script for your platform:

macOS / Linux

curl -fsSL https://storage.tlelabs.com/ctxe/dev/install.sh | sh

Windows (PowerShell)

irm https://storage.tlelabs.com/ctxe/dev/install.ps1 | iex

Verify the installation:

ctxe --version

Index your project

Navigate to your project root and run:

cd ~/projects/my-app
ctxe index .

CtxE will parse your code into an AST, generate vector embeddings, and build a knowledge graph. You'll see output like:

Indexed 1,284 files (4,872 chunks) in 41s

Indexing is incremental — subsequent runs only process changed files.

Your first query

Ask a natural language question about your codebase:

ctxe ask "How does user authentication work?"

CtxE performs multi-round retrieval — semantic search, full-text search, and graph expansion — then returns synthesized evidence with source locations:

Entry points:
  • POST /auth/login — src/api/auth/login.ts:24
  • AuthMiddleware.verify() — src/middleware/auth.ts:56

Core logic:
  AuthService.authenticate() — src/services/auth.ts:89
  - Validates credentials against database
  - Issues JWT access token (15m) + refresh token (7d)
  - Stores session in Redis

Call graph depth: 3 levels | 12 direct files | 28 indirect files

Use with MCP

CtxE runs as an MCP server, so AI tools like Cursor, Claude Code, and Codex can query your codebase context automatically.

The install script automatically configures CtxE as an MCP server for Claude Code — no extra setup needed. Once installed, Claude Code can access your indexed codebase context out of the box.

For other MCP clients (Cursor, Codex, etc.), see MCP Integration.

ctxe mcp install — manual MCP setup command for additional clients — is not available at this time.

Next steps