 |
VOOZH | about |
|
VOOZH | about |
We’re so glad you’re here. You can expect all the best TNS content to arrive Monday through Friday to keep you on top of the news and at the top of your game.
Check your inbox for a confirmation email where you can adjust your preferences and even join additional groups.
Follow TNS on your favorite social media networks.
Become a TNS follower on LinkedIn.
Check out the latest featured and trending stories while you wait for your first TNS newsletter.
This week, Anthropic announced agent view in Claude Code, a CLI dashboard that lets developers manage multiple Claude Code sessions from a single screen.
Where running parallel agents used to require managing multiple terminal tabs and a tmux grid (not to mention the taxing mental load of keeping track of it all), Anthropic’s new product purports to give developers a one-stop shop for Claude Code session management.
With the agent view, developers can launch new agents, send them to the background, and easily jump between different sessions to reply inline or attach a full conversation. Status indicators show which sessions are running, which are waiting for input, and which have produced a pull request (PR).
But while a concise interface is a nice-to-have, it still leaves many of developers’ more pressing concerns unresolved.
Some say yes, some no.
“For engineers that prefer to work in the terminal, agent view does a good job of centralizing the status of running agents threads,” Tom Moor, head of engineering at Linear, tells The New Stack. For him, that’s a step up from chasing information spread across multiple terminal windows.
But when asked whether the agent view will really change the way developers work day-to-day, Rob May, CEO and co-founder at Neurometric AI, tells The New Stack he isn’t convinced yet: “It removes some friction, but it doesn’t change the underlying problem.”
“A better dashboard doesn’t make the agents more reliable. The hard part isn’t visibility. It’s trust.”
He acknowledges the benefit of having a unified interface (“real progress,” he says) but affirms that Anthropic’s new product doesn’t really solve deeper developer issues: “A better dashboard doesn’t make the agents more reliable. The hard part isn’t visibility. It’s trust.”
If Anthropic hopes to steer developers towards more supervisory work, May says getting there will require more than an interface change; namely, developers still need policy-as-code, exception routine, and real audit trails.
If the agent view doesn’t seem immediately dazzling, what kind of help can developers expect from the new product?
Anthropic says agent view can help developers manage long-running agents, calling out “PR babysitters” and “dashboard updaters” as viable use cases.
When asked if developers are really ready to let agents run unattended or semi-attended, Moor and May are both optimistic but cautious. May says it’s “the next logical step for AI-forward companies,” but notes that teams still need to pay attention to processes and results — and ideally hold back on anything that touches production systems.
In general, developers seem ready to hand the reins over to agents, even unattended, for low-risk tasks; but risky tasks that can lead to costly, difficult-to-untangle mistakes require a human in the loop.
“Errors in long-running jobs are expensive to find and fix,” says May. “The debugging burden alone should make developers cautious.”
Lately, that’s always the question on the mind when Anthropic enters the chat.
The AI company has faced a string of complaints in recent months as Claude Code users have reported hitting usage limits faster than usual. But thanks to a recent compute spending spree, during which Anthropic forged deals with SpaceX, Amazon, Google, and Broadcom, and Microsoft and Nvidia, the AI company seems to be actively working at increasing compute capacity quickly.
Claude Code users will have to hope these agreements do the trick, as Anthropic announces “Usual rate limits apply” for agent view.
May doesn’t think developers should turn a blind eye.
“This is one of the most underappreciated problems in agentic development right now,” he says. And as developers run parallel sessions, token costs and rate limits are only going to become problematic faster.
There’s another resource agent view risks exhausting too soon, Moor points out: human mental bandwidth: “You quickly become overloaded, context-switching between several different ongoing agents.”
It’s interesting food for thought. Will agent view end up increasing developers’ workloads, ultimately making them responsible for working on more things all at once?
Finally, it’s worth noting that Anthropic does allow organizations to switch agent view off, which may allow teams to more easily control costs (which can easily spiral out of control when running multiple agents) and help with compliance and auditability.
Dubbing agent view the “one place to manage all your Claude Code sessions,” it seems Anthropic is keen on steering developers toward a supervisory role, where they trust agents to run in the background and just check in when necessary — but developers don’t seem to be convinced just yet.
“It’s a useful piece, but it’s not the control plane developers have been waiting for.”
What’s still missing in the broader agentic development stack is the governance and auditability to drive trust for production use.
May puts it nicely: “Most [enterprises] are stuck in pilot purgatory, not because they lack visibility into their agents, but because they haven’t solved reliability and accountability at scale.” If that’s the case, then agent view may be “a useful piece, but it’s not the control plane developers have been waiting for.”
