The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard
Minor World

Anthropic lets firms run Claude agents in-house

By The Daily Byte ·14 Nov 2026 ·2 min read
Server racks in a data centre server room

Anthropic has opened a private beta of self-hosted Claude Managed Agents, letting businesses run the company's long-running AI agents entirely inside their own cloud or on-premises infrastructure. The announcement was published on 12 November 2026.

Managed Agents is Anthropic's service for autonomous agents that work for hours at a time, fixing broken builds, testing releases, triaging alerts and working through bug backlogs. It first launched as a hosted service in April 2026, with self-hosted sandboxes added in May. Until now, teams that wanted the whole service inside their own environment had to build the platform themselves.

What runs on the customer's servers

With the self-hosted option, every part of the service runs on the customer's side, including the agent harness, sandboxes, credentials and session history. Customers operate the Managed Agents server, its database and the workers where agents run shell and file tools. Session history is stored in the customer's own PostgreSQL database, and developers keep building agents with the same core APIs as the hosted service.

The model itself does not move. Each model call still leaves the customer's site and goes to a provider the customer chooses: Anthropic directly, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry or the customer's own gateway. Anthropic says it has no remote access to a self-hosted server or its logs.

Access is limited for now

The beta is available only to eligible organisations, which must request access through their Anthropic account team. Anthropic has published no pricing for the server software, named no beta customers and given no date for a wider release. Some features of the hosted service are not yet included in the self-hosted version, and the licence published alongside the launch gives customers eight hours to roll back a version after a safety notice and one working day to apply a safety update.

Anthropic says the design keeps credentials out of the sandbox, limits each agent environment to an approved list of hosts, and records every step an agent takes, so activity can be reviewed in the monitoring tools a company already uses.

Sources: Anthropic and Geekbyte.