= 20" />

GROKING

A terminal-first coding agent — multi-provider, streaming, parallel by design
xAI Grok • OpenAI • Anthropic • Live token streaming • Parallel worker agents with 3-way merge

  ██████  ██████   █████   ██  ██  ██  ██   ██   ██████
  ██      ██   ██ ██   ██  ██ ██   ██  ███  ██  ██
  ██  ███ ██████  ██   ██  ████    ██  ██ █ ██  ██  ███
  ██   ██ ██  ██  ██   ██  ██ ██   ██  ██  ███  ██   ██
   ██████ ██   ██  █████   ██  ██  ██  ██   ██   ██████

Why Groking?

Most AI coding tools run in the cloud or inside an IDE. **Groking** runs where you already work — your terminal. It connects directly to xAI's Responses API, gives the model full access to your local filesystem and shell, and supports a real coding loop:

**Inspect → Edit → Run checks → Refine**

• Multi-provider — xAI Grok, OpenAI, and Anthropic Claude behind one CLI (pick with --provider)
• Live token streaming — model output renders as it arrives, not after the spinner finally stops
• Patch-style refactoring — multi-file changes via unified diffs (apply_unified_patch)
• Session continuity — pick up where you left off across terminal sessions, per-provider
• Multi-agent parallelism — a planner breaks complex goals into isolated worker agents that run in parallel; merges use 3-way recovery to survive context drift
• Zero config — just provide a key for your chosen provider and start coding

---

Quick Start

# Install globally (recommended)
npm i -g groking-cli

# Ru