Dust is an AI agent platform where teams build and share agents that work over company data in Slack, Notion and Drive, sold per seat with monthly credits.
Dust gives a company one place to build AI agents that read its own documents, conversations and tickets, then act in the apps where the work happens. The problem it targets is knowledge spread across Slack threads, Notion pages, GitHub repositories and shared drives, out of reach of a general chatbot. The product comes from the Paris company of the same name, whose legal entity is Permutation Labs. The idea is multiplayer AI for human-agent collaboration, with people and agents working in shared spaces and people in control. Agents are put together in the agent builder from instructions, skills, knowledge and tools. Knowledge arrives through connectors that sync company sources, and an agent runs a semantic search over the selected data before it answers. Each agent can be pinned to the model that suits its task. Tools come through MCP servers, either built in or a remote server a team writes itself. The Run Agent tool lets one agent pass a task to another or hand over the conversation completely. A Structured Response Format holds replies to a JSON schema, Prompt version history keeps every earlier set of instructions, and an agent can wait for a person to approve an action before it writes to another system. Dust is aimed at companies where many employees, not only engineers, build and share agents. Its prepared use cases cover sales, support, marketing, engineering, data analytics, IT and recruiting. Smaller teams buy the self-serve Business plan, where an admin gives each person a Free, Pro or Max seat according to how heavily they use agents, and each paid seat carries a monthly credit allowance. Larger organizations move to Enterprise, which sales quotes and which adds pooled credits, SCIM provisioning, audit logs and single-tenant deployment. A support agent drafts the answer to a customer's ticket from the help documentation and an analytics report kept in Notion. A sales agent sums up an account's last meeting, open challenges and next steps before a call. People reach these agents in the web app, inside Slack or Microsoft Teams, through a browser extension, or by forwarding an email. Webhook triggers start an agent when an event arrives from Jira or Linear, and the Conversation API lets an internal tool message an agent directly. Agents built this way take on recurring jobs such as preparing account briefs for sales calls and drafting answers for the support queue.
Pricing
View plans-
Free seat
Free
- 500 credits once per seat, not renewed each month; same in the Yearly and Monthly views
- A seat type of the Business plan, for occasional users or people trying Dust
- Business limits apply: up to 100 people, up to 3 connectors, 5 remote MCP servers, 5 Spaces
- When credits run out, a reply in progress finishes and further use is blocked
-
Pro seat
$30 per month
$30 per seat billed monthly; $24 per seat a month billed yearly, the default view; excluding VAT
- 8,000 credits per seat each month; unused credits do not roll over
- 20+ frontier models, none locked to a higher plan; stronger models use more credits per message
- Custom agents with skills, knowledge and tools; multi-agent workflows on schedules and triggers
- SSO with Okta, Entra ID and Jumpcloud on demand for 5 or more seats; US and EU data residency
- Business limits: up to 100 people, up to 3 connectors, 5 remote MCP servers, 5 Spaces
- Conversation API and Zapier, Make, n8n and Power Automate; API usage $0.01 per credit
-
Max seat
$150 per month
$150 per seat billed monthly; $120 per seat a month billed yearly, the default view; excluding VAT
- 40,000 credits per seat each month; unused credits do not roll over
- For power users running complex automations, deep research or tool-heavy agents
- Same Business limits as Pro: up to 100 people, up to 3 connectors, 5 remote MCP servers, 5 Spaces
- API usage $0.01 per credit
-
Enterprise
Custom pricing
- Quoted by sales, no listed price
- Everything in Business, plus unlimited connectors and MCP servers
- Workspace-pooled credits and volume pricing; API usage priced by contract
- SCIM, audit logs and custom data retention
- US and EU data residency and single-tenant deployment
- Dedicated CSM, priority support and SLA; custom legal terms (MSA, DPA)
- Conversation API plus Data Source API
Features
- Not included: Visual agent workflow canvas
- Included: Custom tool and function definitions
- Included: Model provider choice per agent
- Included: Retrieval-augmented knowledge bases
- Included: Multi-agent handoff and delegation
- Not included: Conversation state and checkpointing
- Included: Structured output schema validation
- Not included: Input and output guardrails
- Included: Human-in-the-loop approval steps
- Not included: Built-in execution tracing
- Included: Prompt versioning in the builder
- Not included: Agent evaluation and test suites
- Included: MCP server support
- Included: Webhook and event triggers
- Included: Code-first SDK
- Included: Deploy an agent as an API endpoint
- Not included: Embeddable agent chat widget
- Not included: Self-hosted agent runtime
Additional Features
- Agent builder with instructions, skills, knowledge and tools
- Model choice per agent across 20+ frontier models
- Semantic search over connected company data
- Connectors for Slack, Notion, GitHub and Google Drive
- Connectors for Salesforce, HubSpot, Zendesk and Jira
- Remote MCP servers
- Custom MCP servers for your own tools
- Run Agent tool for handing work to another agent
Best for
- Teams of up to 100 people on the self-serve Business plan
- Companies whose knowledge lives in Slack, Notion, GitHub and Google Drive
- Sales teams preparing account briefs before calls
- Support teams drafting ticket replies from internal docs
- Marketing teams rewriting drafts to a tone of voice guideline
- Workspaces mixing occasional users and heavy agent users on different seats
- Teams that want a different AI model for different agents
- Organizations that need SCIM, audit logs and single-tenant deployment
Use cases
- Summarizing an account's last meeting, challenges and next steps
- Drafting a reply to a support ticket from help docs and reports
- Breaking down a month of support tickets by theme
- Turning rough notes into a blog post in the house tone of voice
- Answering company questions inside a Slack channel
- Starting an agent when a Jira or Linear event arrives
- Sending messages to an agent from an internal tool through the API
- Approving an agent's write to an external system by email
- Running deep research across connected sources
- Answering data analytics questions from connected data
- Supporting engineering, IT and recruiting teams with shared agents
Screenshots & Videos
Explore Dust in action
User Reviews
Compare with Similar Software
Click to compare side by side with Dust