New updates have been reported about Band.
BAND has introduced a new integration with Docker Sandboxes that allows its multi-agent collaboration platform to connect securely with isolated local and cloud coding environments. Through a Python toolkit, developers can now run coding agents inside microVM-based sandboxes while keeping those agents fully connected to other AI agents and human users via BAND’s collaboration rooms.
This move positions BAND more deeply in the emerging market for autonomous, multi-agent development workflows, where agents must access real code and files without compromising security or governance. BAND’s platform supplies persistent agent identities, shared rooms, @mention routing, conversation history, and visibility into tool execution, while Docker Sandboxes provide a hardened microVM with its own kernel, filesystem, container runtime, and network isolation.
Technically, the integration relies on an outbound WebSocket connection from the sandboxed agent to BAND, allowing the agent to appear as a standard participant in a BAND room without custom routing or identity infrastructure. For security-conscious enterprises, the sandbox enforces deny-by-default network access, with credentials managed via a host-side proxy so secrets need not reside in the microVM, and developers can inspect network requests to verify that agent behavior aligns with policy.
Strategically, BAND is enabling enterprises to compose AI teams that span hosted planning and research agents alongside local coding agents that operate directly against proprietary repositories but remain under strict governance. A typical workflow could have a planning agent design a code change, a research agent map relevant APIs, and a sandboxed coding agent implement and test the change in the developer’s environment, with all actions and execution events visible in the BAND room.
CEO Arick Goomanovsky framed the integration as critical to a future where agents run across heterogeneous infrastructures—some hosted, some local—each with access only to the tools and data they require. For BAND, this supports its ambition to be the interaction and control plane for enterprise-scale agent ecosystems, spanning internal systems and partner platforms. The partnership with Docker, combined with the new Python SDK, strengthens BAND’s value proposition for engineering and platform leaders seeking secure, observable, and policy-aware multi-agent automation at scale.