Axel
Todoist for AI coding agents
330 followers
Todoist for AI coding agents
330 followers
Axel helps you run AI agents and keep them fed. Queue up work, dispatch to the right agent, and approve or deny actions from one inbox. It's native macOS, keyboard-driven, and works with Claude, Codex, OpenCode, and Antigravity out of the box. We hope it helps you ship faster 🚀
Humans in the Loop
@Axel is a beautifully-crafted task manager for AI coding agents. Native for macOS and built for speed (read: keyboard-driven), Axel works with Claude, Codex, OpenCode, and Antigravity out of the box.
If you're working with multiple coding agents, this might be your new favorite tool.
Oh and it's open-source: View source code
Happy shipping!
Axel
Thank you @fmerian 🙏
Hey PH — I'm Ludo, the author of Axel, and I'm a todo list junkie. I've literally tried them all.
Coding agents are redefining my day-to-day work. I have a few years of devtool experience, so I started exploring a developer experience where my interactions are led by a list of tasks — tasks that can be chained, prioritized, and processed by agents I choose.
Not an agent that happens to have a task list. A task list that happens to command agents.
The workflow I was after: queue up what needs to happen, pick which agents handle what, parallelize across git worktrees so nothing collides, and keep one calm interface while the machine goes to war underneath.
That's Axel — a native macOS app that gives you the calm of a task list while your agents are in war mode. Spinning up another agent in its own worktree feels as cheap as opening a new tab.
The app is built with SwiftUI, inspired by Things. Under the hood it leans on GhosttyKit, tmux, and OpenTelemetry to run and monitor the agents being spun up.
Axel is early but it's already part of my daily flow, and that's usually when I know something's worth sharing. I'd love honest feedback, good or bad! Tell me what clicks and what doesn't.
Humans in the Loop
@ludovic lfg! keep up the great work
@fmerian @ludovic Chaining tasks to chosen agents feels right. We’ve seen teams stay sane by routing different steps to different models via GTWY, so worktrees don’t step on each other.
JDoodle.ai
One of the hurdles I have is waiting for the task to finish to start next or work on two different projects. This seems to solve that. Good work.
Axel
Humans in the Loop
@gokuljd spot on! curious: what are your preferred coding agents?
Axel
been running Claude Code and Cursor side by side for months, biggest pain is when they both want to touch the same file. the worktree approach is smart - does each agent get its own isolated git state or do they somehow sync changes back? also curious how you're handling permissions, like can agents accidentally blow away each others work or is there some kind of conflict detection built in?
Axel
Seems interesting. What is a typical workflow with Axel? Would I be creating multiple tasks which are then spread over multiple agents? Or is it more a 1:1 workflow where I'm "baby sitting" an agent?
Humans in the Loop
Exactly. How @Axel works:
You create a task in the app.
Axel spawns a terminal session via the CLI.
The agent runs in tmux, optionally inside a new git worktree.
The app shows live output, events, and permission prompts.
When the work is done, you close the loop and ship.
Migma AI
Using git worktrees for agent isolation is genius - makes spawning agents "as cheap as opening a tab."
How do you handle the merge back to main when two agents touch related files? Does OpenTelemetry help detect potential conflicts before merge?
Love the Things-inspired UI approach!
Humans in the Loop
@ludovic frame this!
Rainglow Color Themes
Hey! I like this idea. Just wanted to mention that I can't grab the app on my work laptop as the SSL for the site is broken and marked as dangerous. Thought I'd raise it in case it's impacting others too.
Axel
@dayle_rees Interesting, thanks for the heads up. it may have been a flash quirk with vercel - it's working fine.
Whoa, Axel looks incredible! Todoist for AI agents is such a smart idea. Im especially curious how the agent dispatching works with different coding languages. Can it prioritize tasks based on agent specialty?