One agent is a tool. Ten agents are an operation. An operation needs oversight — or it overheats.
The headline in one sentence
Anthropic has rolled out a new view in Claude Code: Agent View. A list that bundles every parallel-running agent session in one place — with status, input needs, and access to the transcript. To open it, you press the left arrow key inside a session or type claude agents into the terminal. Currently a research preview for all paid plans (OnlineMarketing.de, May 12, 2026).
Sounds like a quality-of-life update. But it's a far more interesting statement: even the vendor building the industry's smartest agents notices — without oversight, the whole construct implodes.
Why this mini-feature matters
A single agent is trivial to steer. You see the console, you see the answer, done. But the moment three, five, ten agents work in parallel — one researching, one writing, one testing, one escalating — the picture tips over:
- Which session is waiting on your reply?
- Which one has been idling for two hours because a prompt was unclear?
- Where's the pull request the coding agent announced 40 minutes ago?
- Which agent is hallucinating its way into a dead end right now, with nobody noticing?
Anthropic is now solving this for its own platform with a list. Nothing more. But that list is exactly the admission that parallel agents need a cockpit — not another tab.
What this has to do with your mid-sized business
Anyone who thinks of agents as a "toy for developers" is missing the real shift. The mechanics Anthropic is now mapping for coding workflows are identical to what happens at your company the moment you automate more than one workflow:
- Outreach agent writes 200 emails a day.
- Recruiting agent screens 30 applications.
- Sales agent qualifies inbound leads.
- Support agent answers tier-1 tickets.
- Research agent delivers market briefings.
Five agents. Five inboxes. Five statuses. Without a cockpit, that lands in the same hell you wanted out of: a human has to look everywhere, a human loses the overview, a human makes bad decisions because the context is missing.
A classic tool executes. An agent decides. Multiple agents need someone to coordinate the decisions.
What Agent View solves for Anthropic — and what it doesn't
What it does
- An overview of all open sessions.
- Status display (input required, running, done).
- Inline replies — you don't have to jump into every session.
- Lets sessions keep running in the background.
- Access to the full transcript.
What it doesn't do
- It prioritizes nothing. You see every session, but not which one deserves your attention first.
- It doesn't escalate on its own — for an investor email, a €50,000 contract, an unhappy customer.
- It doesn't cross a tool boundary. If your agent works in Slack, another in your CRM, and a third in your recruiting tool — Anthropic's Agent View helps you exactly none.
The list is a cockpit for a single platform. Useful, but not what you need from a business standpoint.
What you need instead
A cockpit that sees across platforms. A layer that doesn't just know which session is running, but:
- Which agent is doing which task for which account.
- What an action costs — in euros, per lead, per hour freed up.
- When a human must step in (thresholds, triggers, escalations).
- When a human could step in (quality, tone, edge cases).
- Which workflows are connected — and where the handoff stalls.
That's the difference between a developer tool that makes sessions visible, and an agent infrastructure a company can actually run on. Anthropic is building the first. We build the second — GDPR-compliant, hosted in the EU, in a German brand voice, with escalations your spreadsheet understands.
What to take away today
If you don't have a single productive agent yet: don't start with ten. Start with the one workflow that eats four hours of your day. Measure, categorize, automate. In that order.
If you already have three or more agents running, across different tools, with no shared view: you don't have an agent problem. You have a cockpit problem. And that gets more expensive than the agents themselves the moment one of them silently runs the wrong direction.
Effort × Automation = Freedom — but only as long as you know what your agents are doing. The moment you no longer know, the formula tips over.
We'll work through this with you — your workflows, your triggers, your escalations, your cost per hour freed up. At the end you'll know whether you need one agent today, five, or first the cockpit, before the next one joins.
Before others do.