Evlat

Product Introduction

1. Definition: Evlat is a native macOS menu bar application and system-level process monitor designed specifically for developers and engineers using AI coding agents like Anthropic's Claude Code and Codex. It functions as a lightweight, persistent visual dashboard.
2. Core Value Proposition: Evlat exists to solve the context-switching problem and lack of visibility when running multiple, long-duration AI agent sessions or background shell commands. Its core value is providing immediate, at-a-glance status on which processes are active, waiting for user input, or have completed, directly from the edge of the screen without requiring window focus or permissions.

Main Features

1. Visual Status Bar & Session Rings: Evlat docks a thin, customizable strip (left or right screen edge) that displays one animated ring per active session. Rings pulse amber when an AI agent (Claude Code/Codex) is waiting for a permission or answer. The visual coding (amber for waiting, other states for progress/done/failed) allows for instant peripheral awareness. The technology uses native SwiftUI animations that halt completely when no activity is present to conserve resources.
2. Universal Command Monitoring via evlat watch: The included command-line tool can wrap any shell command (e.g., evlat watch ./long_running_script.sh). It passes through all terminal I/O, colors, and signals (like Ctrl-C) while relaying the command's progress and state to the Evlat UI. This extends monitoring beyond AI sessions to any long-running script, build process, or deployment task.
3. Programmatic State Control via evlat signal: For integration into custom scripts and workflows, the evlat signal command allows precise programmatic control. Developers can send flags like --progress 0.75, --waiting, --done, or --failed to update the corresponding ring in the Evlat interface. This enables custom jobs, data pipelines, or local servers to report their status visually.
4. Remote Server Monitoring with SSH Tunneling: Evlat can monitor processes on remote servers. After adding a server in the "Remote machines..." menu, it establishes an ssh -R (reverse SSH) tunnel. The remote machine must have the evlat command installed, allowing jobs and Claude sessions on that server to appear in the local Evlat bar, tagged with the remote hostname. This is critical for monitoring cloud instances, CI/CD runners, or staging servers.
5. Non-Intrusive Interaction & Quick Access: Hovering over the bar expands it to show session details (name, duration, task). Clicking a "waiting" ring slides out a card with the exact command the agent wants to run and a "Go to session" button that brings the owning terminal or IDE (VS Code, Orca) to the foreground. The mascot icon acts as a quick launcher for a Claude Code chat balloon (triggered by click or ⇧⌘Space) that runs locally without stealing application focus.

Problems Solved

1. Pain Point: Lost productivity due to constant tab/window switching to check on the status of multiple, parallel AI coding agent sessions or background shell jobs. The "out of sight, out of mind" problem leads to agents sitting idle, waiting for user input.
2. Target Audience: The primary user personas are software engineers, DevOps professionals, data scientists, and technical researchers who extensively use AI-powered coding assistants (Claude Code, Codex) for development, debugging, and automation. Secondary users are system administrators managing long-running processes on local and remote machines.
3. Use Cases:
   • A developer running three simultaneous Claude Code sessions for refactoring, testing, and documentation, needing to see instantly which one has hit a permission prompt.
   • A data scientist running multiple Python training scripts with evlat watch, monitoring their progress from the screen edge without blocking the terminal.
   • A platform engineer tracking the status of deployment scripts across several remote staging servers from a single, unified local dashboard.
   • Integrating evlat signal into a custom CI/CD script to provide a real-time visual progress indicator for team members.

Unique Advantages

1. Differentiation: Unlike generic system monitors (Activity Monitor, htop), Evlat is purpose-built for the workflow of AI-assisted development. Compared to checking terminal tabs manually or using basic terminal multiplexers, it provides a unified, persistent, and visually intuitive status layer. It requires no permissions (Accessibility, Screen Recording) unlike other UI automation tools, as it uses a local socket (127.0.0.1:48151) and CLI hooks.
2. Key Innovation: The combination of a zero-permission, native macOS menu bar app with a versatile CLI tool that works for both wrapped commands (watch) and direct integration (signal). The SSH tunneling for remote monitoring is a significant technical innovation, bringing remote process visibility into a local, lightweight UI. Its commitment to efficiency—drawing zero frames when idle, using native Swift/AppKit—ensures it has minimal system impact.

Frequently Asked Questions (FAQ)

1. Is Evlat secure and private? Yes. Evlat is a local-first application that asks for no macOS permissions (Accessibility, Screen Recording, etc.). All communication happens via a local loopback network interface (127.0.0.1), and it connects directly to your local Claude Code installation. No data is sent to external servers.
2. What are the system requirements for Evlat? Evlat requires macOS 14 Sonoma or later and a Mac with Apple silicon (M-series chips). It is a native ARM64 application, notarized and signed by Apple for security. The file size is approximately 3 MB.
3. Can I use Evlat with OpenAI's ChatGPT or GitHub Copilot? No, Evlat is specifically designed to integrate with Anthropic's Claude Code and Codex agents via installed hooks. Its core functionality is tied to the interaction model and permission prompts of these specific AI tools.
4. How does Evlat monitor remote servers? You add a server's SSH details in Evlat's menu. It sets up a secure reverse SSH tunnel (ssh -R). You then install the evlat command-line tool on the remote server. Processes monitored by the remote evlat command report their status back through the tunnel to your local Evlat application.
5. Does Evlat affect the performance of the commands it watches? No. The evlat watch command is a lightweight wrapper that forks the process. It does not interfere with the STDIN/STDOUT/STDERR streams or the execution of the command itself; it only relays state changes.