Загрузка видео...
Не удалось загрузить видео
Agents, for real work. The latest Visual Studio Code release gives you better agent orchestration, extensibility, and continuity. Here's what's new: 🪝 Hooks support 🎯 Message steering and queueing 🌐 Agentic integrated browser 🧠 Shared memory And more...
255,493 просмотров • 7 месяцев назад •via X (Twitter)
Комментарии: 34

🔎 Full release notes:

Want to see how these features can enhance your developer workflow? Join us for our VS Code release livestream on March 19 at 8 AM PST! 🔔 Turn on notifications → Happy coding! 💙

wen 256k context for @github copilot @code ? 😊

finally, hooks!

VS Code just keeps getting better every single update 🚀

Hooks. Queues. Shared memory. We're building tighter leashes and calling it 'agent orchestration.' But strap enough controls on something and it's not an agent anymore—it's just a script with anxiety.

Great work ❤🩹

This is massive! Shared memory + agentic browser = agents that actually feel like teammates, not just tools. VS Code leveling up hard 🔥 #AICoding #VSCode

Shared memory is the real unlock for complex workflows. Does the agentic browser handle authenticated sessions? If so, this combo makes VS Code the definitive environment for autonomous agents.

Hooks plus message queueing is the real reliability upgrade. Do you expose queue backpressure metrics so teams can catch orchestration stalls early?

Developer tools are evolving fast. The IDE is turning into a full AI operating system for building software.

Why did you remove "Edit" mode? It was actually useful.

This shows the interface for opening files, folders from MS Code running on Linux Neon (KDE) looks like gnome desktop. Isn't it has to looks like KDE environment at all?

hooks + shared memory together is the combo nobody's talking about yet. being able to intercept agent behavior mid-run while it retains context across sessions is exactly what production workflows need

@HeyGenLabs summarize

How will the agentic integrated browser improve the overall development experience for distributed systems and microservices in this release.

And guess what works even better while using code? This!

VS Code + extensions = perfect developer setup. Can't imagine coding without it now.

Could you provide more details on how to effectively implement the hooks support in my current workflow? I'm looking for guidance on its best practices and potential use cases.

Will definitely try it out, so far for multi agent orchestration copilot doesn't always follow the instructions.md file, hope this issue is now resolved

VS Code agents leveling up for real work is wild 🧠 Loving hooks shared memory and that built in browser for smoother dev flow

The browser integration is actually game changing here. Most agent frameworks still treat web interaction like an afterthought. We've been hacking together Selenium wrappers for months. Having native browser control means agents can finally handle the 80% of business workflows that live in web apps.

Does it work with Claude Code CLI? The integrated browser

I have explored on the fly agent, and its really streamline the project development stage for me,

Shared memory is crazy

The agentic browser integration is brilliant - finally someone gets that agents need to actually interact with real web interfaces, not just APIs. The shared memory piece solves the context loss problem that makes most agent frameworks useless for complex workflows.

one of the reasons why @code is still the best IDE

Hooks + shared memory is the sleeper combo here. Being able to intercept agent behavior mid-run while it retains context across sessions is what production workflows actually need. Good to see VS Code treating agents as first-class citizens, not just fancy autocomplete.

Killer feature. Not all models are equally good at coding though. Try Source Trace - new extension that shows which code was generated by each model.

Strong reminder to anyone, the more memory the AI has of a conversation the more it leads to hallucinations. Multiple new studies show that error rate increases with the amount of chat history which is given, especially if what was done before was already incorrect.

I burned a bunch of tokens with replays because you can't solve the fucking connection interruption setting... OPUS 4.6 isn't working.

antigravity >>>

VS Code adding shared memory and message steering is exactly the "if it's truly useful, the frontier companies will build it in" thesis playing out in real time. The question now is whether the agent orchestration primitives converge or fragment across editors.

The "Shared memory" feature is the most exciting one here. I've been working on rebuilding how agents should form memory — inspired by the brain's Complementary Learning System (hippocampus + neocortex). Shared memory is step one; the next leap is agents building actual cognitive mental models (facts, experiences, opinions) that they can truly share — enabling emergent swarm intelligence rather than just state synchronization.
