AI Briefing
KO

Anthropic's Conway Test Set to End — Broader Rollout Possible

·2026.07.21 09:00

Key point

Anthropic's internal always-on agent experiment Conway will end on July 24, raising questions about whether it will be fully discontinued or transition into a full launch.

Details

Anthropic's internal always-on agent experiment Conway will end at 5 PM PT on July 24. Conway is a standalone agent first unveiled this spring, equipped with the ability to run Claude Code, operate browsers, send notifications, and auto-execute via webhooks, and it has been tested by employees and a select group of acquaintances.

The end-of-test announcement can be interpreted in two ways. First, it could mean Anthropic has solidified its strategy around Cowork, which runs on users' local devices, instead of remote-container-based agents. Second, it could signal that internal testing has concluded and the product is moving into a broader rollout phase, such as a limited preview for Max subscribers.

In the latter case, Conway is expected to be integrated across Anthropic's apps as a remote container that allows plugin/skill installation, task delegation, and webhook configuration. It is also expected to come with a feature, via the UI tabs standard, that lets users define and share their own component panels.

This decision will serve as a litmus test for which side Anthropic takes in the industry-wide debate over whether persistent agents should reside on-device or in the cloud.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.