Journeys

Setup & Onboarding

Go from a freshly installed binary to a fully connected setup: LeanCTX auto-detects every AI editor on your machine, writes their MCP config and a shell hook, then verifies the connection — usually in a single command.

You just installed the lean-ctx binary and nothing is wired up yet. This journey covers every command that connects LeanCTX to your editors and your shell — and exactly what each one does.

Quick path

For most users, setup is three commands:

# 1. Install
curl -fsSL https://leanctx.com/install.sh | sh

# 2. Auto-configure everything
lean-ctx onboard

# 3. Restart your shell
source ~/.zshrc

That’s it. onboard auto-detects every AI tool on your machine and writes the MCP config for each one.

What “being set up” means

For LeanCTX to work, three things must be true:

  1. Your AI tool knows about LeanCTX — its MCP config lists the lean-ctx server
  2. Your shell knows about LeanCTX — a hook compresses command output
  3. A data directory exists~/.lean-ctx/ holds stats, sessions, and config

Available setup commands

CommandWhen to use
lean-ctx onboardFirst-time setup, zero questions
lean-ctx setupFull 12-step wizard with all options
lean-ctx install --repairFix drift without prompts
lean-ctx bootstrapCI/automation (non-interactive)
lean-ctx init --agent <name>Wire up one specific editor
lean-ctx doctorDiagnose connection issues
lean-ctx statusQuick “am I connected?” check

Verify your setup

After onboarding, confirm everything is connected:

lean-ctx status     # quick "am I connected?" check
lean-ctx doctor     # deep diagnosis, with fixes for any drift

status shows which editors are wired up and whether the daemon is running. If anything is off, doctor explains the cause and lean-ctx doctor --fix repairs it.

One editor at a time

If you only want to connect a single tool — or add one later — use init:

lean-ctx init --agent cursor      # or: claude, codex, opencode, pi, gemini, windsurf, zed …

Each agent’s exact MCP config, shell hook and prompt snippet are generated for you in the per-editor setup on the Getting Started page.

Beyond coding

Once connected, your agent gets more than compressed file reads. The Web & Research layer lets it pull web pages, PDFs and YouTube transcripts into context as compressed, cited evidence — no extra setup required.

Full documentation

The complete setup guide — including platform-specific installation, manual editor configs, troubleshooting, and the detailed 12-step reference — lives on the Getting Started page.