
Axel is an accelerated task manager for Mac specifically designed for developers who work with multiple AI coding agents such as Claude, Codex, OpenCode, and Antigravity. It provides a centralized task queue where users can add jobs, assign them to the most suitable agent, and monitor their execution in parallel. The core value of Axel lies in unifying the fragmented workflows of different AI agents into a single, keyboard-driven interface that reduces context switching. By eliminating the need to switch between separate agent tools and approval systems, Axel helps maintain focus and accelerates development velocity. The app is natively built with SwiftUI for macOS, with support for iOS and visionOS, allowing users to keep tabs on their tasks from any Apple device. This cross-platform capability makes it a versatile tool for developers on the go.
Developers using multiple AI agents often encounter pain points: each agent comes with its own approval system, skill configuration, and user interface. This fragmentation leads to constant context switching, wasted time, and increased mental overhead. Manual management of permissions for file edits, command executions, and API calls across agents can become chaotic. Axel directly solves these problems by providing a single inbox where all agent requests are visible. Users can see full context—file paths, diff previews, and command arguments—before approving or denying. Additionally, Axel standardizes skill management via portable skill files, ensuring agents are consistently configured without duplication. This unified approach saves time and reduces errors.
The Task Queue is the foundational feature for organizing work in Axel. Users can add tasks to a queue, then select which agent should handle each task—for example, Claude for code generation, Codex for testing. Tasks can be reordered by dragging, and priorities are adjustable even while tasks are running, without requiring a restart. Execution happens in parallel across agents, maximizing throughput. Keyboard shortcuts are available for every action: new pane, dispatch, reorder, and kill, allowing users to maintain a mouse-free workflow. This feature significantly boosts productivity by batching work and letting developers trust that the right agent will execute each assignment at the right time. The queue works with any supported agent, giving flexibility.
Portable Skills address the challenge of replicating agent configurations across different projects. Users define skill files in `~/.config/axel/skills`, and when an agent is launched, Axel symlinks these skills to the agent’s expected location. This ensures that a skill written once works identically with Claude, Codex, OpenCode, or Antigravity. A concrete example is the command `axel -w feat/auth`, which spawns a git worktree and a tmux session for that branch. An AXEL.md file defines the layout, including panes, skills, and grid positions via YAML frontmatter. Sessions persist even after closing the terminal, enabling developers to reattach later without losing context. This portability simplifies sharing and version control of agent workflows.
admin
The Inbox feature centralizes all permission requests from agents. Every file edit, command run, or API call requires explicit approval unless auto-approve rules are configured. The inbox displays full context: the exact file path, a diff preview, and the full command arguments. Users can approve or deny each request individually, or set rules to automatically approve read-only operations or small edits under a token threshold. macOS notifications alert users when an agent is blocked waiting for approval, ensuring timely responses. This system balances security with convenience, allowing routine tasks to proceed automatically while keeping sensitive operations under human control. The auto-approve rules are configurable and apply across all agents.
Axel’s overall workflow starts with adding tasks to the queue, either manually or via the `axel` command line. Each task is assigned to a specific agent based on the user’s choice. As tasks execute, any actions that require permissions are routed to the Inbox for approval. Users can monitor progress, reorder tasks, and adjust priorities in real time. The project’s AXEL.md file ensures consistent environments across sessions by defining panes, skills, and layouts. Token and cost tracking provide visibility into resource usage per task and cumulative totals for the session. This integrated approach transforms a collection of independent AI agents into a cohesive, manageable system that streamlines development from task creation through execution and approval.
Concrete use cases demonstrate Axel’s utility. A developer working on a new feature can run `axel -w feat/auth` to create a worktree and tmux session, then queue tasks for Claude to write authentication middleware and Codex to execute tests—all managed from a single interface. Another scenario involves a team that uses different agents for code review and deployment; Axel’s unified inbox ensures all requests are approved or denied from one pane, with full diffs and logs. Users can set auto-approve rules for harmless operations like reading package.json, while edits to critical files require manual approval. These workflows reduce overhead and accelerate development cycles, enabling teams to focus on higher-level decisions. The ability to reorder tasks on the fly also helps adapt to changing priorities.
Axel is built for developers on macOS who actively use AI coding agents and need a streamlined way to orchestrate them. It is particularly valuable for individual contributors and small teams managing multiple agent sessions daily. The native SwiftUI app integrates with the macOS menu bar, supports Spotlight search, and offers comprehensive keyboard shortcuts for efficiency. While pricing details are not specified, the app is available for free download from the official website. With support for iOS and visionOS, Axel extends beyond the desktop to mobile monitoring. In summary, Axel provides a much-needed unified interface for the emerging multi-agent workflow, transforming scattered AI tools into an accelerated task management system that boosts productivity and reduces friction.
Software developers and AI engineers on macOS who work with multiple AI coding agents such as Claude, Codex, OpenCode, and Antigravity. Ideal for power users who need to orchestrate concurrent agent tasks, manage granular permissions, and automate git workflows. Also beneficial for small teams adopting multi-agent development pipelines who require a unified interface for task management and approval.