Vincent Reboul

Hello Dust team! I was looking to verify the usage of API calls to Anthropic for Claude models and to OpenAI. I am very surprised to see no activity for the API keys used in Dust. However, the models respond normally when I use Dust. Is there someone who could contact me or a support team available to clarify this issue? Thank you.

Hello ! This morning I ofter see this error using an agent with Claude 4 Sonnet. It never happened before. Is there something wrong with Dust and Anthropic?

Hey Dust Team! Are there any plans for a Dust mobile app? It would be amazing to take our agents on the go, interact with them seamlessly, and even use voice instead of typing prompts manually. That would really unlock a new level of flexibility and accessibility!

Hi everyone, thanks for the MCP server addition feature - it's a complete game-changer for using Dust agents. I have a suggestion for an improvement that would greatly enhance security in MCP server management. For certain external servers, we can have a set of tools that allow reading, updating, writing objects, etc. We can currently configure confirmation requests and action management in "Tool stake settings" to increase security. It would be great if we could also block an agent from using a specific tool - for example, to prevent accidentally deleting a CRM contact. I realize that creating guardrails through agent text prompting isn't 100% reliable, so it would be much more effective and secure to be able to restrict tool access at the MCP server settings level. What do you think?