LmCast :: Stay tuned in

AgentsDock: An IDE designed for agentic AI research

Recorded: Sept. 13, 2026, 2:09 a.m.

Original Summarized

AgentsDock - A dock for all your agents

AgentsDock

Home
Docs
Discord

GitHub
·
Open source
·

Beta testing now

AgentsDock
An IDE designed for agentic AI research. AgentsDock currently supports Claude Code, Codex, and Cursor in one desktop and mobile workspace.

Download for macOS


Download for iPhone & iPad

>_
Download for Linux · x86_64

>_
Download for Linux · ARM64

Download for Windows

Download for Android

Trusted by researchers at

How it works

1Set up your server with a few simple commands
2Access the server from any device
3Have your agents do research and model training for you

Read the full setup guide →

We support:

Claude Code

Codex

Cursor

Tailscale (optional)
tmux (optional)

bash
# 1 · Install the AgentsDock server
$ git clone https://github.com/ZhengyiLuo/AgentsServer.git
$ cd AgentsServer
$ ./install.sh

# 2 · Launch the app and paste your server URL
$ open -a AgentsDock

💻 Built for AI researchers

A dock for all your agents
Chat with different agents — Claude Code, Codex, and more — side by side in one place.

Your servers & devices

Georgia's MacBook

100.100.100.11

Georgia's Mac mini

self host server

100.100.100.12

lab-workstation

self host server

100.100.100.15

Georgia's iPad

100.100.100.13

Georgia's iPhone

100.100.100.14

Connect multiple servers, from any device
Your lab workstation, your Mac mini at home, a rented GPU box — connect them all and switch between them anytime, from every device you own.

View and edit code
Open, edit, and syntax-highlight files on your remote — no separate editor needed.

Full terminal access from anywhere
Attach to your chat's persistent tmux session from any device — the same shell, everywhere.

Simulation eval complete: humanoid loco-manipulation. Checkpoint ckpt_0421_best passed the paper-readiness sweep.

92.4%success
0.18 mfinal error
42ksteps

loco_manip_rollout.mp4 · ckpt_0421_best

Check video job results
Agents return plots, images, and rendered rollouts inline — review a run's real output right in the chat.

Loading current release
Bring your agents into the Dock.
Install the desktop client, connect it to your AgentsDock server, and keep Claude and Codex available from every machine you use.
See all release versions →

⌘macOSUniversal · macOS 14+
Download .dmg

>_LinuxAppImage · x86_64 or ARM64

x86_64
ARM64

Windowsx64 · Windows 10+
Download .exe

iPhone & iPadiOS & iPadOS 17+ · TestFlight
Download on TestFlight

Androidarm64 · Beta
Download APK

AgentsDock
A dock for all your agents.© 2026 AgentsDock

Docs
Support
Report an issue
Privacy
Discord ↗

AgentsDock is presented as an Integrated Development Environment designed specifically for agentic AI research, functioning as a central dock for managing and interacting with various AI agents. It consolidates access to different agent capabilities, including Claude Code, Codex, and Cursor, into a unified desktop and mobile workspace.

The core functionality of AgentsDock centers on enabling users to manage and leverage multiple agents simultaneously. This capability is achieved by providing a centralized hub where users can engage in side-by-side conversations with distinct agents. The system is architected to support complex workflows where agents perform research and model training tasks.

The setup process for AgentsDock involves establishing a server through a few initial commands, accessing this server from any device, and then utilizing the setup to allow agents to execute research and model training on the user's behalf. The system supports integration with technologies such as Claude Code, Codex, and Cursor, and optionally supports network tools like Tailscale for secure access and tmux for persistent terminal sessions.

AgentsDock is built to facilitate advanced connectivity among disparate computing resources. It allows users to link multiple servers and devices, such as personal workstations, self-hosted servers, or rented GPU boxes, enabling seamless switching and access from any device owned by the user. Furthermore, the platform provides remote code editing capabilities, letting users open, modify, and syntax-highlight files on remote machines without needing separate editors. Full terminal access is provided across these connections, allowing users to attach to persistent tmux sessions regardless of their physical location.

A significant feature is the ability to visualize agent execution results directly within the chat interface. Agents are capable of returning inline outputs such as plots, images, and rendered rollouts, allowing researchers to review the real output of a run directly within the conversation. For instance, the system can report simulation evaluations, such as a humanoid locomotion checkpoint, providing success rates and specific error metrics alongside associated media outputs.

Availability of the platform is broad, supporting various operating systems and mobile platforms, including macOS, iOS, iPadOS, Linux (both x86_64 and ARM64), and Windows. This multiplatform design ensures that the agentic workspace is accessible across the researcher's entire technological ecosystem.