Dust

Dust is a workspace where teams connect company data sources (Slack, Notion, Google Drive, GitHub, Salesforce, Zendesk, Snowflake, and 20+ others via native connectors or MCP) and build custom agents on top of 20+ frontier models (GPT, Claude, Gemini, Mistral, DeepSeek) without writing code. Agents can be chained into multi-agent workflows, run on schedules or triggers, and shared inside persistent team workspaces the company calls "Pods." It positions itself against both general chat tools (ChatGPT) and individual copilots — the pitch is that isolated single-user AI use doesn't compound, so it builds for team-level, cross-functional agent workflows instead.

Unlike many young AI tools, Dust has verifiable, named customer detail rather than generic logos. Clay's operations team reports a Dust-based sales-enablement agent live within 24 hours of adoption and cites specific, attributable metrics: 100% adoption within its 20-person GTM engineering team and roughly 58 hours/month saved (source: Dust's own customer story, self-reported by the customer, not independently audited). Vanta's GTM team says it evaluated seven AI platforms before choosing Dust, and estimates ~400 hours/week saved on QBR prep across 200 reps. Persona reports 80%+ employee adoption and ~300 agents built after a single engineer's weekend prototype went viral internally. These are vendor-published case studies with named, quotable people attached — worth more than an anonymous testimonial, but still self-selected success stories rather than independent research, so treat the specific hour/percentage figures as customer-reported, not verified.

Published tiers (dust.tt/pricing, checked 2026-08-23): Free ($0, 500 credits, one-time lifetime allocation, no workspace pool access), Pro ($24/seat/month, 8,000 credits/month), Max ($120/seat/month, 40,000 credits/month), and Enterprise (custom, pooled/volume credits, single-tenant option, SCIM, audit logs, SLA). All prices exclude VAT. Credits are consumed by both token usage and per-tool-action charges (0 credits for basic orchestration, 1 credit for actions like running an agent or web search, 3 credits for anything touching a data source, file generation, or a third-party MCP integration like Slack/Salesforce/Notion). On the Business plan, once a seat's monthly credits run out the user draws from a shared workspace pool that admins must pre-purchase as top-up packages — there's no pay-as-you-go on Business, only on Enterprise. Programmatic usage (API calls, Zapier, n8n, Slack bots) draws only from the workspace pool, never from individual seat allowances. Net effect: the seat price is not the ceiling on spend for any workspace running knowledge-heavy agents or automations at real volume — those draw down shared credits fast, and a workspace that under-provisions the pool will see users blocked mid-month.

Dust is SOC 2 Type II certified and GDPR compliant, publishes a Vanta-backed Trust Center (trust.dust.com) with downloadable compliance documents, and states data is not used to train third-party models. It offers US or EU data residency and, on Enterprise, single-tenant deployment. It also states it "enables HIPAA compliance" — that phrasing (rather than a HIPAA certification claim) suggests the safeguards are available but compliance still depends on how the customer configures the workspace, which is worth confirming directly with Dust for a regulated deployment rather than assuming turnkey coverage.

Founded February 2023 by Gabriel Hubert (CEO) and Stanislas Polu (CTO), based in San Francisco with a Paris presence. Funding history, cross-checked across Dust's own posts and third-party trackers: €5M seed led by Sequoia (June 2023), $16M additional round led by Sequoia (announced June 2024, passing $1M ARR at the time), and a $40M Series B led by Abstract and Sequoia with Snowflake Ventures and Datadog participating (announced May 18, 2026) — bringing total disclosed funding to roughly $61.5M. At the Series B, Dust said it had more than 3,000 organizations using the platform globally and more than 300,000 agents deployed cumulatively (vendor figures, not independently audited). This is a well-capitalized, Sequoia-backed team three years in, which lowers near-term shutdown risk relative to an unfunded or bootstrapped competitor, though it doesn't guarantee continuity — three-year-old Series B companies still have a normal path to acquisition or pivot.

Dust's core platform repo (github.com/dust-tt/dust) is public and MIT-licensed, actively developed (pushed to same day as this check), with 1,442 stars and 333 forks. This is the actual product codebase, not a docs/examples repo — but the product itself ships as Dust's hosted SaaS; the public repo does not mean self-hosting is a supported or documented path for customers, and there is no self-hosting guide in Dust's own documentation.

No independent, non-vendor review source (G2, Capterra) was readable during this research — both returned bot-detection walls to direct fetch, proxy, and headless browser. So there is no third-party aggregate rating to report here; a buyer should check G2/Capterra directly before deciding, since those numbers move and we could not confirm one as of 2026-08-23. Customer-reported time-savings figures (hours/month, adoption percentages) throughout this listing come from Dust's own published case studies and are attributed to named individuals, but are not independently audited.