Dust (Workspace Agents for Teams)

Dust is an AI agent platform and collaborative workspace designed for teams that want company-grade assistants deeply integrated with their existing stack. It connects to tools like Slack, Google Drive, Notion, GitHub, Zendesk, and external websites so agents can read, search, and reason over the same knowledge your team uses every day. On top of this semantic knowledge layer, Dust lets you orchestrate leading language models from OpenAI, Anthropic, Google, Mistral, and others, so you are not locked into a single provider. The core workflow in Dust is to create custom agents that understand your company context and can execute actions, not just answer questions. Teams can define instructions, attach tools (including custom "Dust Apps" and Model Context Protocol integrations), and wire agents into Slack or the web workspace so they participate as equal co-contributors in channels and projects. These agents can summarize conversations, route tickets, query internal APIs, run SQL, or transform and move data across systems, turning best practices into reusable skills that improve over time. Dust is built for multiplayer collaboration: any teammate can create or adjust agents, and departments share a common environment where knowledge, tools, and notifications are aligned. This makes it particularly suited for operations, support, sales, and engineering teams that need reliable governance, security, and observability over how AI is used. Enterprise features include SOC 2 Type II compliance, GDPR alignment, SSO, SCIM user provisioning, regional hosting options, and zero data retention at model providers. For deployment and pricing, Dust offers a Pro plan targeted at small teams and startups and an Enterprise plan for larger organizations. Pro provides access to advanced models, custom agents with actions, key connections, and native integrations such as Slack and a Chrome extension, with predictable per-seat billing. Enterprise customers get multiple workspaces, SSO, expanded storage, and higher-touch support, making Dust a central AI layer where teams can standardize on agents for internal Q&A, workflow automation, and data-driven operations.

• ✓Route and auto-draft responses for support tickets by connecting Dust agents to Zendesk, Slack, and internal FAQs
• ✓Answer employee questions about policies, processes, and product details by querying Notion and Google Drive through a company agent
• ✓Let sales and success teams generate tailored customer briefs by pulling from CRM notes, email summaries, and documentation
• ✓Have engineering and ops teams trigger internal APIs or SQL queries via agents to run diagnostics or fetch operational metrics
• ✓Use agents in Slack to summarize long threads, extract action items, and push tasks into project management tools

Start by connecting core knowledge sources such as Slack, Notion, Google Drive, GitHub, and Zendesk so Dust can build a semantic layer over your company data. Then create a new agent in the product workspace, write clear instructions describing its role and boundaries, and attach relevant tools or Dust Apps (for example, semantic search, SQL, or custom API actions) to enable it to act, not just answer. For Slack-centric teams, install the Dust Slackbot and invite the agent into key channels where it can summarize threads, answer questions, and trigger workflows; keep prompts concise and reference channels or documents explicitly to improve relevance. Use the model selection controls to choose an appropriate LLM (e.g., a faster, cheaper model for high-volume tasks and a stronger model for complex reasoning), and monitor performance via the workspace to iterate on instructions and tool configurations. For larger rollouts, coordinate with admins to set up SSO/SCIM, define workspaces per department, and establish governance rules around which agents can access which data sources to avoid over-exposure of sensitive information.

There is no permanent free tier, only a time-limited Pro trial, so ongoing use requires paid seats. Pricing is per user, which can become expensive for large organizations that want widespread access but only light usage per seat. As an orchestration layer over external LLMs, Dust inherits limitations from those models, such as occasional hallucinations, sensitivity to prompt phrasing, and variable performance across domains. Effective use depends heavily on correct configuration of data connections and access controls; misconfigured permissions or incomplete data sources can lead to incorrect or partial answers. Advanced actions and integrations may require engineering effort to build and maintain Dust Apps or MCP tools, and non-technical teams might need support to set up more complex automations. Because it is a relatively new platform, teams may need to invest time in change management and training so that agents are adopted consistently across departments.