OpenChamber introduces team rules for AI agents in enterprise mode
Heise Gruppe GmbH & Co. KG has released OpenChamber 2.0.4, introducing an enterprise mode that allows administrators to control AI coding agent model providers and extension installation sources.

Heise Gruppe GmbH & Co. KG has launched OpenChamber version 2.0.4, introducing a new enterprise mode. This mode allows system administrators to gain granular control over which model providers AI coding agents can utilize and from which repositories extensions can be installed.
OpenChamber serves as an open-source interface for the OpenCode AI coding agent, enabling developers to initiate agent sessions, monitor their progress, and review code modifications across various platforms. The new enterprise mode can be activated by an administrator through the OPENCHAMBER_ENTERPRISE_MODE=1 setting. In this mode, users cannot disable it themselves, and all communication with the AI agent will exclusively use the configured model providers.
The enterprise mode also imposes restrictions on extension installations. Only extensions capable of sending data externally can be installed from administrator-approved repositories. Furthermore, version 2.0.4 limits network access for extensions, allowing connections only to HTTPS addresses authorized during the installation process.
The release also rectifies a previous issue where the Jev classification model could transmit data to OpenCode Zen without explicit user selection. Jev is now disabled by default until a classification provider is chosen. Users also have the option to configure their own endpoint for the Jev model.
Additionally, OpenChamber 2.0.4 introduces new features such as session goal assessment and enhanced image handling. Excalidraw drawings are now opened via a separate extension, and zoom and pan functionalities have been added for images.