dust.tt

What Is It

Dust positions itself as “Multiplayer AI for human-agent collaboration” — an AI Agent platform for enterprise teams focused on human-agent collaboration. It lets teams create custom agents that understand company context and deploy them across sales, customer support, marketing, engineering, data, IT, legal, HR, and other functions. The site says it is used by 3,000+ organizations with 300,000+ agents deployed, though concrete customer outcomes are mainly presented through case-study quotes.

Core Capabilities

Dust is not just a standalone chatbot. Its core value is connecting AI agents to enterprise knowledge and tool stacks. Teams can start from prebuilt templates and configure instructions in natural language, while more advanced teams can build custom connectors and agentic tools such as semantic search, SQL queries, data visualization, and multi-step workflows. On the model side, the page explicitly mentions support for choosing among OpenAI, Anthropic, Gemini, Mistral, and others, with the ability to switch when better models become available, reducing the risk of being locked into a single model provider.

Integrations, Security, and Chinese Support

Integrations are one of Dust’s strengths. The captured content lists 59 integrations, covering Slack, Microsoft Teams, Notion, Google Drive, Salesforce, HubSpot, Gong, Intercom, Zendesk, GitHub, Jira, BigQuery, Snowflake, Power BI, and more, with support for Web Crawler and internal API connections. On security, Dust highlights SOC 2, HIPAA, GDPR, control over data selection and hosting location, trusted model selection, fine-grained permissions, and authentication. Chinese support is not explicitly stated, but customer support use cases can draft messages in 50+ languages, and marketing workflows support cross-language translation and localization.

Pricing and Trial

The site offers a 14-day free trial with no credit card required, along with Contact sales and Pricing entry points. However, the captured body text does not disclose specific plans, seat pricing, usage limits, or how model calls are billed. Before purchasing, enterprises should further confirm the total cost, whether model fees are included, and whether integrations and security features are gated by plan tier.

Pros, Cons, and Who It’s For

Dust’s strengths include strong enterprise-context capabilities, broad integrations, multi-model flexibility, and a balance between no-code templates and developer extensibility. It is well suited to “AI Operators” and business teams that want to embed AI into real workflows. Its limitations are that results depend heavily on data governance, permission setup, and continuous iteration; it may be too heavy for individual users or simple Q&A scenarios. Public materials also lack independent benchmark-style evaluation metrics.

Access from China and Alternatives

The captured text does not specify network accessibility from mainland China, payment methods, or local compliance support, so these remain unknown. For teams in China considering deployment, it is necessary to test access stability, SSO/permissions, cross-border data handling, and procurement/payment workflows. Comparable options include Glean, Microsoft Copilot Studio, ChatGPT Enterprise, Notion AI, Salesforce Einstein, HubSpot AI, as well as domestic products such as Dify, Coze, and 扣子企业版.