We pulled vendor documentation, user reports from Reddit and Product Hunt, and what third-party coverage exists when putting this together. G2 and Capterra have almost nothing on Dust yet. The company launched in 2022 and still sits outside the main review aggregators, which makes independent research harder than it should be.
The core idea is collaborative AI, not personal AI. Shared agents, shared knowledge, humans and AI working in the same workspace rather than each person running a private chat thread. That's the bet Dust is making.
What is Dust?
Dust is an AI workspace built for teams. Founded in 2022, headquartered across Paris and San Francisco, the company announced a $40 million Series B in May 2026 and counts more than 3,000 organizations as customers. The product runs a multi-model backend, pulling from GPT and Claude as the primary options, with Gemini and Mistral also available depending on the use case.
What separates Dust from a typical AI chat tool is the multiplayer surface. Agents are shared across departments. Knowledge is centralized. A sales rep, a support manager, and an engineer can all access the same agent and get consistent results because the underlying context is shared, not siloed per user.
Honestly, that's a sharper product idea than most of what's in this category.
The other piece is what they call a semantic knowledge layer. It doesn't just retrieve text from connected data sources. It tries to synthesize meaning across them. Whether that claim fully holds in practice is harder to verify without hands-on testing, but the architecture is more considered than typical retrieval-augmented setups.
Dust Features: Workflows, AI Agents & Automation Capabilities
The agent builder is the center of everything. You create agents scoped to a specific job, attach knowledge sources, give them tools, and deploy them to the right people. Marketing gets a @ContentWriter. Support gets a @TicketRouter. Each one is configured separately but lives in the same shared workspace.
Collaborative workspaces organize teams and agents together with layered permissions. Dust uses RBAC with dual-layer controls, and Enterprise adds SCIM-synced groups for larger org structures. That setup is more mature than most tools in this category bother with.
The integration list covers the usual places teams live. Slack is there. Google Drive and Notion too. On the automation side, they connect to Zapier and Make, as well as native webhook and MCP triggers. Dust claims 100-plus production connectors and separately lists 50-plus native integrations. They also document 12 fully managed, continuously synced knowledge connections, including Google Drive, Notion, Slack, Salesforce, and Microsoft. Those are distinct categories. Worth distinguishing before you assume everything is equivalent.
Fair. That's a solid list for a product this young.
On the workflow side, agents can run on schedules, respond to events, or get triggered via webhooks. Multi-agent orchestration with conditional logic is supported. You can chain agents so the output of one feeds the input of another. Not common at this price point.
Dust Automation Power: How Complex Can Your Workflows Get?
More complex than the homepage suggests. The visual agent builder handles straightforward setups without code. But Dust also exposes a RESTful API, supports MCP servers, and lets developers wire in custom integrations. The ceiling is reasonably high.
Multi-agent orchestration is where the real automation depth sits. Conditional workflows across multiple agents mean you can build something like: scan incoming support tickets, route them by type, escalate edge cases to a human, and log the outcome, all inside Dust without external automation tools.
We cross-referenced their developer docs with user reports from Reddit and Product Hunt. The API is consistently described as clean and well-documented. A few users noted that the initial configuration for complex multi-agent chains takes real time. That's not a knock. Complex setups take time. Good to know going in, though.
Error handling is present. Agents finish in-progress responses before stopping, and admins can set usage controls per workspace. Execution logs run for 365 days on Enterprise. Lower plans get less visibility there. Teams with compliance needs should check plan tiers carefully before committing.
Not alarmed by that. Just flagging it.
Dust AI Agent Capabilities: What Can It Actually Do Autonomously?
Dust's agents aren't chat interfaces dressed up as automation. They can hold context from connected data sources, act on triggers without a human prompt, and hand off to other agents mid-workflow. Genuine autonomy, not a parlor trick.
The platform supports scheduled runs and event-driven triggers. An agent can wake up at a set time, pull data from Notion or a connected CRM, process it, and push output somewhere without anyone clicking anything.
Human-in-the-loop is also built in. The multiplayer surface means a human can step into an agent conversation at any point, correct course, and step back out. That combination of autonomous operation with easy human override is something a lot of multi-agent platforms still handle awkwardly. Dust gets it right.
Product Hunt comments from users in 2023 and 2024 pointed to knowledge synthesis as a genuine differentiator. The complaint that came up more than once was setup time, particularly for teams without a technical admin to configure data connectors properly.
That tracks.
Is Dust Easy to Set Up Without Code?
Relative to the complexity it offers, yes. The agent builder is visual. You don't need to write code to connect Slack, point at a knowledge source, and deploy a working agent. Most non-technical users can get a basic agent running in an afternoon.
The catch is that "basic" and "useful" aren't always the same thing. Getting agents to do something genuinely valuable, like synthesizing across multiple data sources with accurate context, takes more configuration work. A few Reddit threads from 2024 mentioned that teams without an internal AI ops person struggled to get past the surface-level setup.
Dust offers a Chrome extension on top of the web app. No mobile app, no desktop client. Web-only is fine for most teams but worth flagging.
Documentation is strong. They have guides, tutorials, an Academy section, and a developer docs area. The Slack community appears active. Dust has introduced an AI-powered support skill as a first line of contact, and email support is publicly listed, though which plan tier it kicks in on isn't clearly documented. We'd verify that directly with the vendor before assuming coverage at the Pro level.
Dust Pricing: Is It Worth It vs Zapier or Make?
The free plan gives you 500 lifetime credits. Enough to evaluate the product. Not enough to run it in production. Realistically, you're looking at a paid plan quickly if Dust is going to do anything useful for your team.
Pro is listed at €30 per seat per month. Max is $150 per seat per month. Enterprise is custom and requires a sales conversation. The gap between Pro and Max is large. A significant step for teams that need more agent capacity or deeper logging.
Compared to Zapier, which prices on task volume rather than seats, Dust is a different model entirely. Dust is an AI workspace first. Zapier is a workflow automation tool. Not direct substitutes. Teams evaluating both should know that Zapier's pricing scales with usage while Dust's scales with team size.
Refund policy isn't publicly stated. We went through the pricing page and found nothing on it. Not unusual for B2B SaaS, but it's a gap.
We don't love that.
Dust vs Glean: Which Automation Platform Wins?
Glean is the most natural comparison. Both products connect to company data sources and surface knowledge for teams. The difference is what they do with that knowledge.
Glean is primarily a search and discovery tool. You ask it something, it finds the answer. Dust is trying to build agents that act on knowledge, not just surface it.
Dust wins on automation depth. Glean doesn't do multi-agent orchestration or event-driven workflows. Dust does. If the main problem is "I can't find what I need," Glean might be the better fit. If the problem is "I need agents that do things with what they know," Dust has more to offer.
Microsoft Copilot comes up in the same breath. Copilot has obvious distribution advantages inside Microsoft 365 shops. But it's constrained to that ecosystem in ways Dust isn't. Dust connects to Notion, GitHub, and third-party tools that Copilot handles less gracefully.
Guru is closer to a knowledge base tool than an AI agent platform. The comparison doesn't hold up past the surface level.
Who Should Use Dust? (And Who Shouldn't)
Mid-size teams with multiple departments all trying to share the same AI context. That's the fit. A 100-person company where sales, support, and marketing are all building their own private ChatGPT habits and getting inconsistent results. Dust is designed for exactly that problem.
Engineering-led teams that want to go deeper with the API and build custom integrations will get more out of Dust than teams expecting a plug-and-play experience. The product rewards investment. It doesn't reward passivity.
Solo operators and very small teams. They should probably look elsewhere. At $30 per seat, the economics get odd at small team sizes, and the multiplayer premise requires actual team scale to make sense. Something like CrewAI might be worth considering if the primary need is building and deploying multi-agent workflows with more developer control and less team-collaboration overhead.
Companies deep in the Microsoft ecosystem should trial Copilot before committing to Dust. The integration story is meaningfully different.
Dust Review Verdict
Dust is doing something real. The multiplayer AI workspace concept is more than marketing copy. Shared agents, a semantic knowledge layer, and genuine multi-agent orchestration put it ahead of most tools in the category.
The weak spots are predictable for a company this young. No G2 presence yet, so there's no large-scale review signal to cross-reference. Setup complexity is the recurring friction point from every source we found. The jump from Pro to Max in pricing is steep. Support visibility at lower plan tiers needs clarification from the vendor.
None of that disqualifies it. The product is worth evaluating seriously for any team that's already past the "give everyone a ChatGPT account" phase and wondering what comes next.
The $40 million Series B suggests they're building for the long term. The product is already more mature than its age suggests. Worth watching closely.
Frequently Asked Questions
Does Dust work without a technical setup person?
Basic agents can be configured without engineering help, but getting real value from the platform typically requires someone comfortable with admin settings and data connector configuration. Teams without an internal technical resource have reported friction getting past initial setup. The documentation is good, which helps, but it's not a fully self-serve experience for complex deployments.
What AI models does Dust use?
Dust runs a multi-model backend. GPT and Claude are the headline options. Gemini and Mistral are also available depending on the use case, with DeepSeek listed among the supported models as well. You can select models per agent, which gives more flexibility than platforms that lock you into a single provider.
Is there a free plan worth using?
There's a free tier with 500 lifetime credits. It's enough to understand how the product works and build a few test agents. It's not enough to run Dust as an actual team tool. Budget for a paid plan before evaluating it seriously, otherwise the credit ceiling will cut the evaluation short before you've seen what the platform can actually do.
