Product Introduction
1. Definition: Terminal Mode by Even Realities is a specialized ambient computing software application designed for the Even G2 smart glasses. It functions as a lightweight, heads-up interface for monitoring and interacting with command-line-based AI coding agents (such as Claude Code, Cursor, or similar agentic coding tools) running on a host computer. Technically, it bridges a local development environment with an augmented reality (AR) display, transforming the standard terminal session into a live, viewable, and interactive feed on wearable hardware.
2. Core Value Proposition: The core value proposition of Terminal Mode is to enhance developer productivity and maintain workflow ("flow state") by decoupling active coding agent management from the physical laptop. It exists to solve the problem of constant context-switching and screen-checking inherent in managing long-running AI coding tasks. By enabling developers to monitor, approve, and guide their coding agents directly from their smart glasses, it allows for greater mobility and multitasking without losing oversight or control, ultimately maximizing token efficiency and developer throughput.
Main Features
1. Ambient Agent Status Monitoring:
This feature provides a real-time, at-a-glance status dashboard of all connected coding agent sessions directly within the user's line of sight on the Even G2 glasses. It works by streaming critical metadata from the host machine's running even-terminal process to the paired smart glasses via the Even Realities mobile app. Users can visually discern which sessions are active (running), require intervention (blocked), or are idle (awaiting input) without needing to switch windows or unlock their laptop. This leverages the glasses' HUD (Heads-Up Display) capability to present persistent, non-intrusive information.
2. Hands-Free Command and Control: This feature enables two primary interaction methods directly from the glasses' touch-sensitive ring interface: Tap to Approve for simple, pre-defined actions (like file writes or command executions), and Tap-and-Hold to Record Voice Instruction for more complex, open-ended guidance. The implementation involves the CLI tool printing pairing QR codes and the mobile app handling secure authentication and low-latency audio streaming. This turns the glasses into a remote input terminal, eliminating the need to physically return to the keyboard for common agent interactions.
3. Zero-Installer, CLI-First Deployment:
The installation and launch process is streamlined to a single command: npm install -g @evenrealities/even-terminal@latest. This Node.js package runs the server process directly on the host machine, capturing terminal output and agent state. The system uses a QR code pairing mechanism via the Even Realities app to securely link the host to the specific Even G2 glasses. This design prioritizes developer ergonomics, requiring no complex GUI installers and integrating seamlessly into existing terminal-based workflows.
Problems Solved
1. Pain Point: Constant Context-Switching and Screen Fatigue. Terminal Mode directly addresses the disruptive workflow where developers must repeatedly check their laptop screen to monitor AI agent progress. This constant switching breaks concentration ("flow state") and leads to mental fatigue. The product provides a persistent, ambient status feed, solving the "is my agent stuck?" problem by making status visible passively.
2. Target Audience: The primary target audience includes AI-augmented developers ("Token-Maxers"), Vibe Coders, full-stack engineers, and technical leads who rely on multiple concurrent AI coding sessions. This encompasses users of tools like Claude Code, GitHub Copilot Workspace, Aider, or any agentic framework where the AI operates in an autonomous loop requiring human-in-the-loop checkpoints. They are professionals seeking to optimize their output per hour and maintain coding momentum.
3. Use Cases:
- Multitasking During Long Builds/Tests: A developer can start a long-running agent task (e.g., refactoring a large codebase), step away for a coffee or a walk, and still receive critical alerts and approval requests on their glasses.
- Managing Multiple Agents Simultaneously: An architect can oversee the parallel work of three separate coding agents—each on a different microservice—from a single view, tapping to approve simple merges or providing voice notes for complex directions.
- Physical Mobility in the Office/Home: Users can move between a standing desk, a couch, and the kitchen while remaining fully connected to their active coding sessions, approving actions and responding to agent queries without being tethered to their desk.
Unique Advantages
1. Differentiation from Traditional Monitoring: Unlike monitoring via a second monitor, a phone notification, or remote desktop apps, Terminal Mode offers a true ambient, hands-free, and glanceable interface. It is not a full remote desktop but a focused, high-signal dashboard. Compared to standard terminal multiplexers (like tmux), it adds a spatial, wearable dimension that frees the user's physical posture and primary screen, keeping the main display available for other work.
2. Key Innovation: The Agentic-Ambient Interface Loop. The key innovation is the creation of a tight feedback loop specifically optimized for human-AI agent collaboration. It uses the smart glasses not as a general-purpose computer but as a dedicated control surface for autonomous processes. The combination of a lightweight CLI backend, mobile-app-mediated secure pairing, and wearable HUD display creates a new interaction paradigm where the human operator can oversee and direct digital workers from the physical world, effectively treating AI agents as remote colleagues whose status is visible in your peripheral vision.
Frequently Asked Questions (FAQ)
1. What smart glasses are compatible with Terminal Mode, and is an installer required?
Terminal Mode is currently designed exclusively for the Even G2 smart glasses. The installation is minimal and CLI-based: you run npm install -g @evenrealities/even-terminal@latest on your computer (the host). No GUI installer is needed on the host machine; pairing is accomplished by scanning a QR code with the Even Realities mobile app.
2. How does Terminal Mode handle security and prevent unauthorized access to my coding sessions? Security is managed through a QR code pairing process initiated from your authenticated Even Realities mobile app. The initial pairing generates a secure key exchanged between your host machine and your specific pair of Even G2 glasses. All subsequent streaming and command data are encrypted and scoped to that paired session, ensuring that only your glasses can view and interact with your terminal sessions.
3. Can Terminal Mode work with any coding agent or only specific ones like Claude Code?
Terminal Mode is provider-agnostic in principle. The even-terminal server captures standard terminal output and state, so it can work with any command-line-based agentic tool that runs in a terminal session. During setup, you use the --provider flag to specify your agent (e.g., claude, aider, cursor). The glasses will then display the status and allow interactions based on the agent's standard terminal I/O.
4. What is the battery impact of using Terminal Mode on the Even G2 glasses? Using Terminal Mode involves continuous streaming of status data to the glasses' display. While the Even G2 glasses are optimized for all-day wear, streaming dynamic content will consume more battery than using them solely for notifications. Actual impact depends on session length and update frequency, but it is designed for practical use during a workday. The product's stability and UX are frequently highlighted in user reviews, suggesting efficient performance.
5. How does "tap-and-hold" voice guidance work, and what happens to the audio?
When you tap and hold the touch ring on your Even G2 glasses, the microphone activates and records your voice instruction. This audio is captured by the Even Realities app, transcribed or passed as raw audio to the even-terminal server on your host machine, and injected into the active agent's session as a user input or system message. This allows for nuanced, conversational guidance to steer an agent without typing.
