Difficulty in accessing and managing multiple web applications securely while using an AI agent for coding tasks

Signals over time

3 observed signals across 1 sources, tracked for 4 days. Confidence: low.

What people are saying Observed

• “Over the last few weeks I've really been using Codex and Claude to solve a lot of parallel tasks, across a lot of different projects. This lead me to wishing that I could interface with my agents in the same way that I interface with my already existing windows and editors. For me, this usually means that I have like 5 to anywhere around 8 different windows that I drag around depending on where I want to focus, this could mean I am focusing on the terminal now, or I want to test stuff in a simulator or browser; but when I go to interface with agents, I have to manage many tabs, or I have to look through a list in their native apps. This is not convenient for me, and I'm almost certain this is not convenient for many of you. That's why I built an infinite canvas, where my traditional desktop style workflow can come to life alongside my agents, I can drag them around depending on where I want my focus to be, or I can let them just pile up, like the 15 finder windows you're hiding with your browser right now. But, when I do want to go back to that thread, I will intuitively just know right where it is. I didn't stop there though, I really enjoyed the inbox sty”

  Hacker News · frustration

• “> IMO, an agentic operating system should be not only about "integrating the agent everywhere" but also about providing the compartimentalization and tools to allow working safely with agents when they can reach into every part of the system. It should allow to easily extend the system safely. Right, but that's also a contradiction. An "agentic" OS is most interesting to the average AI enthusiast if it can, essentially, be your executive assistant: shop for you, book travel for you, manage your agenda, sort through your emails, do your taxes, and so on - possibly including anticipating your needs before you articulate them. That requires human-equivalent access, more or less. Most people also don't need it - really, most lives are not that complicated - but the allure of your own computerized EA is undeniable. It's your personal servant, how cool is that. The safe and principled way to use agents is to scope permissions appropriately, restrict them to narrowly-defined tasks, etc, but that also makes the technology a lot more boring. And these silos are very difficult to build when you're interacting with third-party services anyway. You co”

  Hacker News · wishlist

• “Let Cursor’s agent use your signed-in Chrome in background tabs. Cursor’s agent can work on your code, but pages behind your logins are harder to reach. That matters when you want it to check a dashboard, pull a report, fill a form in an internal tool, or review a staging site with your real logins. Chrome Relay lets the agent use your everyday Chrome, signed in as you, in background tabs. Cursor runs commands from its terminal through the open-source agent-browser CLI by Vercel Labs. There are four reasons I’d use this approach: It works with the agent you already use. Chrome Relay supports Cursor, Claude Code, Codex, and Paseo, rather than being tied to one AI company’s extension. It stays in the background. Tabs open inactive in a collapsed “Agents” tab group. No focus switch, nothing moving on screen. The agent can follow the whole task. Popups and “open in new tab” become its own hidden tabs. It clicks through Google sign-in hops when you’re already signed in, and stops to ask when a password, passkey, or 2-step verification is needed. It’s fast because Chrome is already open. About 0.15 seconds per command and about 0.6 seconds to open a page. A browser session looks like thi”

  RSS Feeds · wishlist