Dust

Dust is a model-agnostic enterprise platform for building and sharing AI agents that are connected to company knowledge in Slack, Notion, Google Drive, GitHub, Salesforce and other tools. It is built for operations, sales, support and marketing teams that want non-engineers to create governed agents without every department running its own isolated chatbot.

Dust is a Paris-based enterprise AI agent platform founded in September 2022 by former Stripe colleagues Gabriel Hubert (CEO) and Stanislas Polu, who previously spent three years as a research engineer at OpenAI. The company operates as Permutation Labs SAS and positions its product as "multiplayer AI": a shared workspace in which humans and agents work inside the same projects, drawing on a common context layer rather than each employee using a separate assistant. Users can call general agents such as @dust for internal knowledge search or model-specific agents, and teams create their own specialised agents with instructions, skills, knowledge sources and tools, including multi-agent workflows and MCP server connections. The platform is model-agnostic, offering more than 20 frontier models from OpenAI, Anthropic, Google, Mistral and DeepSeek, and connects to Slack, Notion, GitHub, Google Drive, Salesforce and other data platforms with semantic search across them. Governance features include private spaces with role-restricted access, user/builder/admin roles, SSO, audit logs and SCIM on Enterprise, EU or US data residency, and a policy that customer data is never used to train models. In May 2026 Dust raised a $40 million Series B co-led by Sequoia and Abstract, with Snowflake Ventures and Datadog participating, taking total funding above $60 million. At that point it reported more than 3,000 customer organisations, over 300,000 agents deployed, 70% weekly active usage and zero churn in 2025; named customers include Clay, Persona, Profound and Doctolib. Much of the platform code is public on GitHub under an MIT licence. It competes with Glean, Microsoft Copilot Studio, Relevance AI and Lindy.

Category

AI Agents & Orchestration

Pricing

Subscription, Freemium, Contact for pricing

Target Market

CIOs, COOs, Heads of AI, Revenue Operations, Customer Support Leaders

Deployment

Cloud-only, API-based

Founded

2022

Headquarters

Paris, France

Customers

3,000+ organisations (May 2026)

• ✓
  Custom agent builder

  No-code creation of agents with instructions, skills, knowledge sources and tools, so business teams can ship their own agents without engineering help.

• ✓
  Shared multiplayer workspaces

  Humans and agents collaborate in the same projects with shared tools and context, so agents built by one team can be reused across the company.

• ✓
  Connected context layer

  Semantic search across Slack, Notion, GitHub, Google Drive, Salesforce and other sources, plus MCP server connections, grounding answers in current company data.

• ✓
  Model-agnostic runtime

  Access to 20+ frontier models from OpenAI, Anthropic, Google, Mistral and DeepSeek, letting teams pick per-agent models and avoid single-vendor lock-in.

• ✓
  Multi-agent workflows

  Multiple agents can be invoked in one conversation or chained into workflows, handling multi-step tasks such as research followed by drafting.

• ✓
  Governance and access control

  Private spaces, user/builder/admin roles, SSO, and on Enterprise SCIM, audit logs and custom data retention, giving IT control over what agents can see.

• ✓
  Self-improvement loops

  Agents incorporate feedback and memory from how people use them, which Dust says improves agent output over time without manual re-prompting.

• •
  Internal knowledge assistant

  Employees ask @dust questions in Slack or the web app and get cited answers drawn from Notion, Drive and GitHub instead of searching several tools.

• •
  Sales account research

  GTM teams such as Clay's build agents that pull CRM and web context into account briefs and outreach drafts, cutting manual prep time before calls.

• •
  Department-wide agent rollout

  Persona deployed over 300 agents across 11 departments, showing how one governed platform can serve many teams' distinct workflows.

• •
  Company AI strategy at scale

  Doctolib uses Dust as the platform for a company-wide AI strategy covering 3,000 employees, centralising governance while letting teams build locally.

• •
  Support answer drafting

  Support teams build agents that search policies and past tickets to draft consistent replies, reducing handling time and escalations to specialists.

List pricing is published in euros per seat per month, metered as a monthly credit allowance per seat (8,000 on Pro, 40,000 on Max) that does not roll over. SCIM, audit logs, pooled credits, single-tenant deployment and custom legal terms are gated behind custom-priced Enterprise, and third-party reviewers note per-seat costs add up for org-wide rollouts.

This page was written from 7 sources, 4 on domains other than dust.tt.

1. 1.dust.tt — series b multiplayer aivendor
2. 2.dust.tt — pricingvendor
3. 3.dust.tt — securityvendor
4. 4.siliconangle.com — multiplayer ai startup dust swipes 40m funding help enterpri
5. 5.tech.eu — dust raises 40m series b to build the multiplayer operating
6. 6.github.com — dust
7. 7.metaflow.life — dust tt reviews