English | Chinese Original

English translation of the Chinese original. This version is generated for international readers and may be refined over time.

Date: 2026-05-15

Yesterday I found a more interesting website, stock. ai, whose main interface is this.

It's a nice style, different from the traditional AI that makes the front UI, and it's kind of new, kind of pixels.

What kind of website is this?

I log in for a wave.

After login you need add a computer.

Wait, why do you need a computer? And if you want to continue, make sure you have a terminal like claude code/codex/ kimi cli on your computer.

I need to execute this order at the terminal after I add command.

And then I made it through this order.

Linked to prompt me to create an onboard anent.

Because GPT officially sent me six months of Pro 20x, I've managed to get it, so this place has to be full of GPT.

Also, note that both the name and the DEscription here are created before they can be modified.

I don't know how to start the product after it's created, but I'm lucky to have a description of my Agent.

And then I saw this Agent on the left member list.

When I'm finished, I'll probably understand what this product means.

I first understood it as a cli terminal shell, but there's so many. Why do I have to use you?

After the study, I found out that the advantage was that these CLI Agents were equipped with a "Slack Crust".

You can imagine it like this:

So why does it need you when it comes up?

Because Agent was not running into a cloud server in Slock, but on your computer.

Slock is responsible for "how people talk to Agent, how they divide up, how they know each other what they're doing," real code reading, running orders, changing files, or your local Claude Code/Codex.

It's like OpenClaw's access to various IM terminals.

Because if you open it directly in the terminal, Claude Code, it's a single session:

But when using Slock, it becomes:

Directly with Claude Code, you work in the dark room with a programmer.

With Slock, you open an office with a lot of AI colleagues who can be @, who can be seen, who can see the context, who can claim the job, who can discuss it.

This Agent will start on your computer, Codex or Claude Code, read repo, run tests, change the code, and get back to the Slock channel.

If you have another agent, you can make it.

@Agent2 help me review @Agent1 just made changes.

That's when the significance of Slock came out.

It's not about making individual Agent smarter, but about having a shared space of collaboration between multiple Agents and individuals.

You don't have to copy Claude Code's output into the group, or copy their needs back to the terminal. Everyone, including AI Agent, is in the same stream.

I created a couple of Agents, and then I sorted out their personalities and tasks.

Feels a little bit Agent Teams now, and I got excited again.

To the left, the Tasks here, is the task bar, all the assignments are here.

And there's another interesting thing about it: Agent is long-lasting.

Not that you opened the terminal once, it just disappeared. It's that every Agent has his own workspace and memory. It can remember its role, the context of the project, what it has talked about before, and what it is responsible for.

Of course, there's not much to think about here.

So-called memory is essentially a document, a session, something that follows from history. It wasn't a sudden awakening, but it was the Slock that made you a product of the "long-term context."

What's the difference between Claude Code / Codex?

The biggest difference I understand is that direct use of CLI is a tool, and Slock is a collaboration space.

If I just have to change a bug today, open the terminal:

codex

Or:

claude

And then talk to it about need, which is definitely the fastest.

One less system, one less web page and one less synchronized message.

But if your needs begin to turn into this:

• I'd like to have Agent do a different job at the same time.
• I'd like to split a mission into a few different tasks, Agent.
• I want to drop a word on my phone and keep my computer running.
• I want team members to see what's going on with Agent.
• I want Agent to report, discuss, deliver on the channel like a colleague.

That Slock is more natural than the terminal.

To be honest, the end fits "I'm staring at it."

The Slock fits "I throw life out and let it advance in a visible collaboration space."

Is it an introvert? Is it a white prostitute model?

This is not.

The Slock itself does not provide a model scale. It only transfers the CLI tool that you have installed locally.

You use Codex on the bottom, and that's still your OpenAI/ ChatGPT/ Codex interest.

You're using Claude Code on the bottom, and that's still taking your Claude account or Claude Code.

You used Kimi CLI, Gemini CLI, too.

So don't take it as something like "GPT Pro is an API."

It's more like:

Turn your already existing CLI Agent into an AI colleague who can be @ on the web channel.

---

Who do I think it suits?

If you're just one person writing code and you're mostly sitting in front of a computer, just use Claude Code/ Codex.

Slock may be a little heavy for you.

But if you're already using al agent, and you're thinking about it all the time:

Can this mission let a Agent run?
This report?
Can the context of this project be left to a certain Agent for a long time?
Can I just call it work from the web/phone without opening the terminal?

Then Slock is worth trying.

It's not about how to talk to AI, it's about how to work with a group of AI Agents.

That's what it says on the front page:

Where humans and AI agents build together.

Not using AI as an input box, but as a collaborator who can be assigned tasks, can participate in discussions and can be permanently present.

Of course, it looks a little early at this point, and UI has a character and a little ahead of product patterns.

But the direction I'm accepted.

Because when the AI programming tool is getting stronger, the question may not be whether the individual Agent will write code, but:

How do I manage more than one?
How do I know what they're doing?
How do I make them not repeat the same thing?
How do I put humans, code stores, missions, discussions, AI implementation in one place?

Slock wants to do this thing.

Brief summary:

Claude Code / Codex: A strong AI programmer
Slock: an office where many AI programmers work with humans.

If you use it as an alternative to Claude Code, it might not be necessary.

But if you think of it as "Ai Age Slack," it suddenly makes sense.