Boltbar

Product Introduction

1. Definition: Boltbar is a native macOS AI chat bar and developer utility built with Swift and AppKit. It functions as a lightweight, system-wide floating prompt interface designed to integrate large language models directly into any macOS workflow without requiring resource-heavy Electron frameworks.
2. Core Value Proposition: Boltbar exists to eliminate context-switching for knowledge workers and software developers. By allowing users to press ⌥ Space from any active application to query GPT or Claude—or execute autonomous coding tasks via Claude Code and Codex CLI in Agent mode—it streamlines AI assistance into a zero-friction, high-speed macOS productivity layer.

Main Features

1. System-Wide Global Overlay (⌥ Space): Boltbar operates as a native macOS overlay accessible via a global hotkey across all running applications. It features asynchronous response handling, allowing users to queue subsequent messages while a model is generating output, use arrow keys to switch models or thinking effort, and instantly reopen previous chat histories with full-text search capabilities.
2. Autonomous Agent Mode (Claude Code & Codex Integration): Activated via ⌘ J, Agent mode hooks directly into local CLI tools (Claude Code or Codex) to execute multi-step workflows inside designated project folders. It features granular permission controls—Look (read-only), Ask (interactive prompt validation), Edit (unattended file modification), and Full (unrestricted execution)—while automatically parsing project-level instruction files like CLAUDE.md or AGENTS.md.
3. Multi-Modal Support & Secure Local Key Management: The utility supports image, PDF, and text file attachments via drag-and-drop or pasting. API credentials (OpenAI and Anthropic) are stored securely in the macOS Keychain, while local chat history, attachments, and application state are isolated within ~/Library/Application Support/Boltbar to ensure complete user privacy.

Problems Solved

1. Pain Point: High cognitive load and workflow fragmentation caused by constantly alt-tabbing between IDEs, browsers, and standalone web-based or Electron-heavy AI chat applications.
2. Target Audience: Software engineers, full-stack developers, DevOps specialists, system administrators, and technical power users running macOS 13 (Ventura) or later on Apple Silicon or Intel architectures.
3. Use Cases: Rapidly querying technical documentation or debugging snippets without leaving an IDE; delegating localized refactoring and test-fixing tasks to Claude Code or Codex via Agent mode; managing multi-project AI workflows with distinct permission levels and persistent local chat sessions.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Pros: Ultra-lightweight footprint (approx. 12 MB installed, <4 MB download) written in native Swift/AppKit with zero Electron overhead; deeply integrated macOS global shortcut architecture; robust developer-focused features including CLI handoff (/terminal), granular agent permissions, and zero telemetry/analytics privacy guarantees.
   • Cons: Requires macOS 13 Ventura or newer; relies on the user providing their own API keys or having pre-configured Claude Code/Codex CLI logins; advanced features like PDF attachments require active API keys rather than raw CLI logins.
2. Key Alternatives & Differentiation:
   • Vs. Raycast AI / Alfred Workflows: While Raycast and Alfred offer extensible command bars with AI extensions, Boltbar is purpose-built specifically as a deep native AI interaction layer featuring native multi-modal attachment handling, persistent project-scoped chat histories, and direct CLI agent integration (Claude Code/Codex) with safety permission gates.
   • Vs. Standalone Desktop Clients (ChatGPT/Claude native apps): Traditional desktop apps require dedicated window management and context switching. Boltbar provides an instant overlay accessible over any active application with queued message input and local state persistence.
   • Pricing Differentiation: Distributed as a free, open-source MIT-licensed utility installable via Homebrew (brew install --cask juanluisrto/boltbar/boltbar) or DMG, with an optional €19/S19 one-off supporter license that removes occasional reminder screens without locking out core functionality.

Frequently Asked Questions (FAQ)

1. How do I install and update Boltbar on macOS? Boltbar can be installed via Homebrew using the command brew install --cask juanluisrto/boltbar/boltbar. Because it includes an automated in-app updater, standard Homebrew upgrades will ignore it unless the --greedy flag is passed. Alternatively, you can download the signed and notarized DMG or ZIP directly from GitHub or run the shell installation script provided in the official repository.
2. Is my data and API key private? Yes. Boltbar stores your OpenAI and Anthropic API keys securely in the native macOS Keychain. All chat transcripts, history databases, and uploaded attachments remain strictly local to your machine within ~/Library/Application Support/Boltbar. The application communicates exclusively with your chosen AI provider (OpenAI/Anthropic) and GitHub (solely for checking application updates), with zero third-party analytics trackers.
3. What are the permission levels in Boltbar's Agent mode? Agent mode utilizes local Claude Code or Codex CLI tools and offers four distinct permission tiers: Look (restricts the agent to reading files and executing read-only commands), Ask (prompts the user via ⌘ ↵ or ⌥ ⌘ ↵ before executing any command or file modification), Edit (permits unattended file modifications and deletions while prompting before other actions), and Full (unrestricted execution requiring a secondary confirmation toggle). Each project folder remembers its assigned permission state independently.