Dust review

Dust is built for companies where the hard part is not the agent logic but the context. Agents draw on 70+ connectors including Slack, Notion, Google Drive, GitHub, Salesforce and Zendesk, and Dust's pitch is bringing that context together so agents can answer with the full picture and act. Teams build custom agents with skills, knowledge and tools, chain them into multi-agent workflows on schedules or triggers, and pick from 20+ models per task. The permission model is the stand-out: what an agent can access is separated from who can use it, with SCIM-synced groups and audit logs on Enterprise. The self-serve Business plan covers teams of up to 100 with Free, Pro and Max seats, where Pro includes 8,000 credits a month. Prices showed only in euros from here, so the USD figure is pending. Unused credits reset each month.

Who it is for

Companies that want staff across teams to build knowledge-grounded agents with central permissions.

Price

See vendor pricing (USD check pending); free seat with 500 lifetime credits.

US price check pending. See the vendor pricing page.

Key features

• Custom agents with skills, knowledge and tools
• Multi-agent workflows on schedules and triggers
• 20+ frontier and open-source models
• Dual-layer permissions separating what agents can access from who can use them
• SOC 2 Type II, US or EU data residency

Strengths and weaknesses

• Strength: Strong permission model for company data
• Strength: Choice of models per agent
• Strength: SSO on the self-serve Business plan
• Weakness: Prices showed in euros from our location
• Weakness: Unused credits do not roll over
• Weakness: The Business plan caps connectors and remote MCP servers

Key facts

• Seat types: Free, Pro and Max on the Business plan
• Pro allowance: 8,000 credits per seat a month
• Max allowance: 40,000 credits per seat a month
• Business plan size: Teams up to 100 people
• Models: 20+

Where it ranks

Dust is No. 6 in our Best AI Agent Builders verdict.