A terminal interface to higherMark AI, Cerebras 3,000 tok/s, GPT-4o, Claude, Gemini, Grok, and 400+ OpenRouter models. Works in your Mac terminal, Cursor, iTerm2, or Warp.

Installation

1. 1

   Run the one-line installer

   Downloads a standalone binary directly, no Node.js, no npm, no repo clone needed.

   curl -fsSL https://highermark.xyz/install.sh | bashcopy

2. 2

   Verify installation

   highermark /versioncopy

   You should see higherMark CLI v0.1.1.

Quick start

Open your terminal and type highermark to launch an interactive REPL chat session against Cerebras (3,000 tok/s by default).

$ highermark
  higherMark  AI terminal
  ──────────────────────────────
  Model   llama-3.3-70b
  API     https://higher-mark-june-july-prod.vercel.app
  ──────────────────────────────

⚡ llama-3.3-70b › what is the BRL/USD rate today?copy

Pass a message directly for one-shot output (great for scripting):

$ highermark "explain quantum entanglement in one paragraph"copy

Model selection

Specify a model by name after the command. Aliases resolve to full model IDs automatically.

# Use Grok
highermark grok

# Use Claude
highermark claude

# Use GPT-4o
highermark gpt

# One-shot with a model
highermark grok "who won the 2026 World Cup?"

# Switch mid-session with /model
⚡ llama-3.3-70b › /model deepseek
✓ Switched to deepseek/deepseek-r1copy

List all models live from OpenRouter with /models, this fetches the full catalogue in real time.

⚡ llama-3.3-70b › /modelscopy

Chat commands

Usage & credits

Config file

Settings are stored at ~/.highermark/config.json. You can edit it directly:

{
  "apiBase": "https://higher-mark-june-july-prod.vercel.app"
}copy

Set a custom API (e.g. local dev server) with /setapi http://localhost:4000.

Cursor integration

higherMark CLI works natively in the Cursor integrated terminal. Open a terminal in Cursor with Ctrl+` (backtick) and run highermark.

Using the AI terminal alongside the Cursor editor

The CLI uses higherMark's shared pool (Cerebras + OpenRouter keys), not your Cursor subscription. This means AI usage in highermark doesn't count against your Cursor Pro token limit.

Split terminals

Run higherMark AI in a dedicated split pane while keeping your editor and Cursor AI in view.

Cursor

# Open terminal: Ctrl+` (backtick)
# Split vertically: drag the terminal panel divider
# Or use the terminal split icon in the top-right of the terminal panel
highermark grokcopy

iTerm2

# Split pane vertically
⌘+D

# Split pane horizontally
⌘+Shift+D

# Navigate panes
⌘+[ and ⌘+]copy

Warp

# Click the split icon (+) on any pane header
# Or use the command palette: ⌘+P → "Split right"
highermark deepseekcopy

macOS Terminal.app

# New tab: ⌘+T
# Split view (macOS 10.11+): View → Show Split Viewcopy

Piping & scripts

higherMark CLI auto-detects when stdin is piped and produces clean output (no ANSI colors, no prompts).

# Summarize a file
cat README.md | highermark "summarize this in 3 bullets"

# Review code with a specific model
cat src/app.js | highermark claude "find potential bugs"

# Pipe to pbcopy on macOS
highermark "write a commit message for: added CLI tool" | pbcopy

# Use in a shell script
SUMMARY=$(highermark "summarize: $(cat notes.txt)")
echo "$SUMMARY" >> weekly-digest.mdcopy

Available models

The /models command fetches this list live from OpenRouter. Below is a static reference of commonly used models.

⚡ Cerebras (fast inference, default)

OpenRouter models

Run /models in the REPL to see all 400+ models with live pricing from OpenRouter.