Onepilot and Kittylitter both drive coding agents from your phone, but with different ambitions. Kittylitter is a free, open-source agent client that can even run Codex right on the phone, with clever local network, SSH, and peer-to-peer connections to your own machines. Onepilot is a native operations app: it deploys, schedules, and supervises a fleet of AI coding agents on servers you own, with a real terminal and a full dev workflow around them. Kittylitter is centered on running a single agent; Onepilot is centered on managing a fleet.
What is the difference between Onepilot and Kittylitter?
The core difference is single agent versus managed fleet. Kittylitter is an inventive, early open-source client focused on running one agent, sometimes on the phone itself over built-in peer-to-peer connectivity. Onepilot is an operations platform: it stands up persistent agents through a wizard, supervises a fleet across machines, schedules overnight runs, and gives you a real shell and dev tools underneath. Both drive Codex from your phone; only Onepilot deploys agents as standing services, manages many from one view, and runs them on a schedule.
| Capability | Onepilot | Kittylitter |
|---|---|---|
| Drive coding agents from your phone | ✓ | ✓ |
| Claude Code as a first class agent | ✓ | Beta |
| Codex as a first class agent | ✓ | ✓ |
| Deploy a persistent agent with a guided wizard | ✓ | No |
| Manage a fleet across many machines, one view | ✓ | No |
| Multi framework (OpenClaw, Hermes, more) | ✓ | No |
| Cron scheduling: agents run while you sleep | ✓ | No |
| Real terminal, any shell | ✓ | No |
| Run agent on device, no server | No | Codex |
| Edit remote files | ✓ | No |
| Review git diffs in app | ✓ | No |
| Port forward a localhost preview | ✓ | No |
| Peer to peer NAT traversal | Via Tailscale | Built in |
| Skills marketplace for agents | ✓ | No |
| Open source | No | ✓ (GPLv3) |
Can Kittylitter run Codex and Claude Code from your phone?
Kittylitter runs Codex as a first-class agent and can even run it directly on the phone with no server, and it supports Claude Code in beta. Onepilot runs both Claude Code and Codex as first-class agents in a real session on your server. The difference is not whether the agent runs, it is what surrounds it: Onepilot deploys the agent as a persistent service through a wizard and supervises it, while Kittylitter focuses on running the agent itself, on device or over its peer-to-peer link.
Can you deploy, schedule, and manage a fleet of agents with Kittylitter?
No. Kittylitter is built around running a single agent, not deploying and supervising many, and it has no built-in scheduling. Onepilot sets up a persistent agent through a guided wizard (framework, model, keys, channel) across OpenClaw, Hermes, and more, keeps it running and visible on your server, shows every agent on every host in one dashboard, and runs any agent or task on cron so your fleet works while you sleep.
Does Kittylitter give you a real terminal and dev workflow?
No. Kittylitter focuses on the agent itself and does not include a full shell or in-app dev tools. Onepilot adds a real terminal for any command plus a dev ecosystem: edit remote files, review git diffs before committing, and forward a localhost port to preview a dev server. Kittylitter's standout instead is connectivity: built-in peer-to-peer NAT traversal that reaches machines behind firewalls without extra setup, where Onepilot relies on Tailscale for the same reach.
When should you use Onepilot instead of Kittylitter?
Use Onepilot when you want a managed fleet, not a single agent. It fits if you want to deploy, monitor, and schedule agents on your own servers; if you run several machines and want one fleet view; if you want a real terminal with git diffs and localhost preview under the agent; or if you want a framework agnostic setup with overnight, scheduled runs. If running one agent, sometimes on the phone itself, is the whole job, Kittylitter may be enough.
Where does Kittylitter win over Onepilot?
Kittylitter wins on open source and on-device flexibility. It is free and open source under GPLv3, can run Codex directly on the phone with no server at all, ships built-in peer-to-peer connectivity that traverses NATs and firewalls, and is a fun place to be early on a fast-moving project. If free and open source are non negotiable, Codex is your daily driver, or you want to run an agent on the phone itself, Kittylitter is worth a look.
Should you choose Onepilot or Kittylitter?
Pick Onepilot if…
- You want to deploy, monitor, and schedule agents on your own servers.
- You run several machines and want one fleet view.
- You want a real terminal with git diffs and localhost preview under the agent.
- You want a framework agnostic setup and overnight scheduled runs.
Pick Kittylitter if…
- Free and open source are non negotiable.
- Codex is your daily driver and running it on device appeals to you.
- Built-in peer-to-peer connectivity is a priority.
Ready for a managed fleet, not just a single agent? Start now. For the wider picture, see the agent deploy overview and running Claude Code on iPhone.