Dust

What Dust is

Dust is an AI agent builder for business teams. Its homepage calls it "Multiplayer AI for human-agent collaboration": a platform that connects company knowledge, tools, and teams so people can create, share, and run agents across real workflows. The company's own guides describe it as a place where business teams describe agents in plain language and deploy them across the whole organization. The reader it is written for is the operations, marketing, support, or go-to-market team that wants working agents without a ticket to engineering. Its pages do not say where the company is based.

The company's guides make a plain argument for why this matters. The people who understand business workflows best are often not engineers, so a no-code platform lets those domain experts build the solution directly instead of writing specs for someone else to translate. Dust also gives its power users a name. It calls them "AI Operators," people who build AI with judgment, deploy it, and run it for a whole team, and says the platform is being built for them.

What you build and how

The editor is an agent builder driven by plain-language instructions and visual configuration. The company's guide to no-code agent builders breaks the work into four parts. Instructions define behavior. Tools extend what the agent can do. Model selection affects performance. Deployment controls who can use the agent and where it appears. Its walkthrough, titled "How Dust's agent builder works," describes defining the agent's instructions and scope, connecting data sources such as Notion or Salesforce through pre-built connectors, and selecting which capabilities the agent can use. All of it happens through visual interfaces rather than code, and the guide describes a loop of changing a setting and testing immediately.

The data layer is the connector set. The homepage says to connect Dust to 70+ connectors, including Slack, Notion, Google Drive, GitHub, Salesforce, and Zendesk, so agents have the context they need to get work done. It makes a specific point about context: most AI tools can retrieve a message from Slack or a record from your CRM, while Dust brings the relevant context together across Slack, CRM, docs, and other company systems so agents can answer with the full picture and take action.

Models are a choice, not a lock-in. The homepage says you can build agents with 20+ frontier and open-source models from OpenAI, Anthropic, Google, Mistral, and more, and choose the model that fits each workflow.

Logic is mostly instruction-led. The company's guides contrast chatbots with agents. A chatbot retrieves an answer or follows a pre-scripted path. An agent built on a no-code platform can plan multi-step workflows, decide which tools to use from context, and run tasks across several systems. The guide's example is an order-status question: the agent checks the order in the CRM, identifies delays, sends the customer an update, and logs the interaction. That is the guide's illustration of the category, not a Dust customer case.

Collaboration is the product's main idea. The homepage lists "Multiplayer AI for cross-functional work": bringing people and agents into the same workflow with shared context, reusable skills, connected tools, and human review. People and agents share access to company knowledge, connected tools, conversations, and notifications. The company puts it this way: people guide the work while agents handle repeatable tasks across teams.

Deployment is organizational rather than public. The guide says you might deploy a support agent to your entire organization through Slack, or restrict an HR agent to specific teams using space-based access controls. The homepage's use-case list includes keeping user and API docs up to date from code changes. The navigation also points to a Chrome Extension, a Developer Platform, a Platform Documentation page, and a Github Repo, so there is a path into code for engineering teams. Its pages do not describe what those developer surfaces contain.

Governance is where Dust spends a full homepage block, headed "AI-native governance." You set permissions for data, tools, systems, and agents, then use audit logs and analytics to monitor activity, adoption, usage, costs, and ROI. A second block says "Dust's dual-layer permission model separates what agents can access from who can use them," and names SCIM-synced groups.

What the pages say

The pages show no prices. The site's navigation includes a Pricing page and two buttons, "Try for free" and "Request a demo," but no plan names, prices, seat limits, or usage limits appear on its pages. Check the Pricing page before you budget.

Scale claims come from the company. The homepage says "Trusted by teams at 3,000+ global organizations," and its closing section reads "300,000+ Agents deployed" and "3,000+ Teams running on Dust." A row of case-study cards sits under the first claim without naming customers in its text. Two testimonials are named. Everett Berry, Head of GTM Engineering at Clay, is quoted calling Dust "the most impactful software we've adopted since building Clay," and Ryan Wang, CEO at Assembled, says Dust "became the connective tissue that amplifies what each team does best." A table-of-contents entry in the company's guide reads "Watershed scales AI adoption to 90% with Dust," which is the company's own headline for a case study.

The press section lists three items, each with a date. Sifted ran "From chatbots to 'digital teammates': The shift towards multiplayer AI" on Jul 13, 2026. Axios ran "Agentic AI startup Dust raises $40M" on May 18, 2026, and Forbes ran "Why Dust Believes Agentic AI Has To Be A Team Sport" the same day. Those are headlines on the company's own page.

On security, the pages describe the permission model, SCIM-synced groups, audit logs, and a Trust Center link in the footer. They do not name a compliance standard such as SOC 2 on its pages, so look at the Trust Center page for certifications. Support appears as a Slack Community link, a Support link, and a Become a Partner link.

Who it suits and who it does not

Dust suits a 20-person operations, support, or go-to-market team that already works across Slack, Notion, Google Drive, Salesforce, or Zendesk and wants agents that read from those systems. It also suits an IT group that needs permissions and audit logs before it lets many people build agents. The company's own framing, shared work between people and agents, fits teams that treat agents as colleagues with a defined scope rather than as a widget on a website.

It is a weaker fit for a solo founder who wants a public website chatbot. The pages describe agents deployed to an organization through tools like Slack and restricted by team spaces, and they do not describe an embeddable customer-facing widget. The company's guides also say the main work is defining workflows and writing clear instructions, not learning the platform. A team that has not written down how its support queue or approval process works will spend its first weeks doing exactly that.

Two other limits are worth knowing. First, prices are not published on its pages, so a small team cannot work out cost per agent or per seat from its pages. Second, the platform leans on third-party models and on the quality of the connected data, so an agent is only as good as the documentation it can reach. A developer who wants an escape hatch into code has a Developer Platform and a Github Repo listed, but would need to read those pages to see how far they go.