Anthropic tests 'Conway' — AI agents that work autonomously around the clock
What it really says
Anthropic is testing a persistent agent platform called 'Conway' that turns Claude into an always-on environment. Conway can execute code, support external webhooks, interact with Chrome, and send notifications — an 'always-on' AI agent.
Our assessment
AI agents that run permanently represent a paradigm shift: from 'I ask a question and get an answer' to 'AI works independently in the background'. This is unsettling for many — but it's ultimately automation like we know from cron jobs, auto-replies, and notification systems. The difference: the AI agent makes its own decisions. The question is how well controls and boundaries are set.
Relevance for Germany
Relevant for companies wanting to deploy AI agents. McKinsey's 'State of AI Trust 2026' study warns: governance requirements increase significantly with autonomous agents.
Fact check
Conway has only been tested in limited beta access so far. Details come from a combination of Anthropic product pages and tech reporting.
Source
- • Dataconomy DE 03.04.2026
- • McKinsey State of AI Trust 2026