Dust

Overview

Dust is a cloud-based AI platform built for what its official site calls “Multiplayer AI for human-agent collaboration.” It enables teams to move beyond isolated conversations with a single assistant and instead create shared workspaces where humans and multiple AI agents operate together as a collaborative team. The platform is designed for production use: agents can be given distinct roles, instructions, and tool access, then deployed into an environment where their work is visible, interruptible, and auditable. Dust combines the conversational capabilities of large language models with systematic orchestration, making it easier to build AI applications that are both interactive and governed.

At the core of Dust are agents and rooms. An agent is an AI worker with a defined identity and capability set, while a room provides a collaborative context where agents and human team members interact. Dust’s runtime tracks agent activity in real time, exposing status signals such as online, idle, and busy, so collaborators always know whether an agent is available or actively working. The interface renders agent behavior as chat cards, inline tool chips, status indicators, and mention-based interactions, giving every action a clear, transparent trail. This design supports not only human-agent collaboration but also agent-to-agent workflows, where specialized agents can pass tasks and context to one another inside the same scenario.

From an engineering perspective, Dust offers a browser-based environment with built-in observability and control. The dynamic floor view provides a live map of the workspace, showing which agents are active, where conversations are happening, and what tools are being invoked at any moment. The platform also includes privacy-conscious design, such as input masking, to protect sensitive information during shared sessions. Because the entire solution runs in the browser, no desktop installation is needed, and teams can start building or experimenting within minutes. The product emphasizes iterative refinement: agent outputs can be reviewed inline with reactions and textual feedback, and those insights can then be used to update agent instructions, tool configurations, or overall workflow design.

Typical use cases include automating knowledge work, building internal AI assistants, orchestrating multi-agent content pipelines, and creating auditable human-AI operations. Dust is suitable for AI engineers who need a scalable agent orchestration layer, as well as business teams that need a safe, collaborative way to involve AI in daily processes. By making agent behavior observable and collaborative, Dust addresses the trust and coordination challenges that often limit large language model deployments in organizations. In short, Dust positions agents as productive teammates rather than opaque black-box tools.

Core capabilities

Multiplayer agent workspaces

Build and operate teams of AI agents that coexist with humans in shared, persistent rooms.

Real-time presence and status

Agents expose online, idle, and busy states, giving teams live visibility into availability and workload.

Interactive chat cards

Agent outputs appear as rich in-stream cards with text, lists, tool references, and mentions for clear communication.

Mention-based collaboration

Users can @mention specific agents to delegate tasks or ask for input, just like addressing a colleague.

Tool integration chips

Agents can invoke external tools, and each tool usage is displayed as an inline chip in the conversation stream for transparency.

Reaction and review controls

Humans can react to agent messages with emoji-style pills and provide contextual feedback to improve future behavior.

Activity visualization

The dynamic floor view shows active versus idle agents, working states, and conversational flow to support monitoring.

Multi-agent orchestration

Multiple agents can be active in the same scenario, coordinated by the platform runtime to handle complex workflows.

Privacy-conscious design

The interface supports input masking and privacy controls to protect sensitive information in shared sessions.

Browser-based deployment

No local installation is required; teams can access the platform and start collaborating through the web.

Use cases

Automating internal support and FAQ handling

Running multi-agent research and summarization workflows

Delegating data extraction and analysis tasks to AI agents

Orchestrating content creation and editing pipelines

Monitoring and auditing AI agent behavior in production

Simulating collaborative scenarios for training or demonstration

Streamlining cross-functional operations with assisted decision-making