Dust Release Notes

Sound Notification When an Agent Needs Your Approval

What is it?

Dust can now play a sound alert whenever one of your agents pauses and waits for you to approve or decline a tool action. The sound only plays when you're not actively viewing that particular conversation, so you're notified precisely when your attention is needed elsewhere.

Why is it useful?

When an agent runs a task that requires manual approval (for example, before it performs a sensitive action), it pauses until you respond. If you've switched to another tab or you're working on something else, it was easy to miss these moments and leave an agent waiting. This audio alert ensures you can step away or multitask without losing track of pending approvals.

How does it work?

Once enabled, Dust plays your configured sound whenever an agent is waiting for approval or decline of a tool, and only if you're not currently looking at that conversation. Sound notifications are off by default, so you're fully in control of whether you use them.

Concrete Use Cases

Here's how you could use it:

• Multitasking across tabs: You launch an agent on a long task, switch to your inbox, and get an audible cue the moment it needs your go-ahead, no need to keep checking back.
• Running several agents at once: While monitoring multiple conversations, you're alerted as soon as any one of them pauses for approval, so none of them sit idle.

Benefits for you

Less time spent checking on agents, faster response to approval requests, and smoother multitasking. Your agents spend less time waiting, which keeps your workflows moving.

How to access it?

Sound notifications are off by default. To turn them on, go to Personal Settings → Notifications and enable the sound alert (you can also configure which sound plays).

Set Default Skills in Pods

🎯 What is it?

Pod editors can now select a set of default skills directly from their Pod settings. Once configured, these skills automatically appear in the input bar of every new conversation started in that Pod, exactly as if each user had added them manually via the "/" command or the dropdown menu.

💡 Why is it useful?

Until now, if you wanted to use skills that aren't attached to a specific agent, you had to add them by hand every time you started a new conversation. This created repetitive friction, especially for teams relying on the same set of skills day to day. Default skills remove that step, ensuring the right capabilities are always ready from the very first message.

⚙ How does it work?

A Pod editor opens the Pod settings and selects the skills they want available by default. From that point on, every new conversation in the Pod automatically loads those skills into the input bar. Users can still add or remove skills per conversation as needed.

✨ Concrete Use Cases

Here's how you could use it:

• Standardizing team workflows: A Pod dedicated to customer support can have your knowledge-search and ticket-handling skills preloaded, so every agent starts with the same toolkit.
• User onboarding and activation: Set helpful skills as defaults to guide new members and help them get value from the Pod immediately, without needing to know which skills to add.

📈 Benefits for you

You save time by eliminating repetitive setup, ensure consistency across conversations in a Pod, and make it easier for new users to get productive right away with the most relevant skills already in place.

🚀 How to access it?

This is available to all users. If you're a Pod editor, head to your Pod settings, select the skills you'd like to set as defaults, and they'll automatically appear in every new conversation started in that Pod.

🎯 What is it?

You can now choose a default agent for an entire Pod. Once set by a Pod editor, every new conversation started in that Pod will automatically begin with the selected agent, instead of the standard default. This setting lives directly in your Pod settings.

💡 Why is it useful?

Teams often rely on a specific agent that already has the right context, skills, and connected data for their work. Until now, members had to manually pick that agent each time they started a conversation, an easy step to forget and a source of inconsistency across a team. Setting a default agent at the Pod level removes that friction and ensures everyone starts from the same, well-prepared assistant.

⚙️ How does it work?

An editor of the Pod opens the Pod settings and selects the agent they want as the default. From that point on, all new conversations created in the Pod open with this agent already selected.

✨ Concrete Use Cases

Here's how you could use it:

• Support team Pod: Set a customer-support agent as the default so every new conversation is handled by an agent that already knows your product documentation and support tone.
• Sales team Pod: Default to an agent connected to your CRM and sales playbooks, so account managers immediately get relevant, context-aware answers without any setup.

📈 Benefits for you

• Consistency across your team: everyone talks to the same specialized agent by default.
• Time saved: no need to manually select the right agent for each new conversation.
• Better answers from the start: conversations begin with the context, skills, and data your team actually needs.

🚀 How to access it?

This is available to all users. If you're an editor of a Pod, open your Pod settings and select the agent you'd like to set as the default for new conversations. Existing conversations are unaffected.

What is it?

When a trigger fires, the conversation it creates now lives directly inside the Pod you've selected. This means every automated run lands in the right workspace context from the start, rather than sitting outside your team's organized spaces.

Why is it useful?

Automated work can quickly become hard to track when it's scattered outside your team's normal environment. By routing triggered conversations into a Pod, everything stays organized, easy to find, and shareable with the right people — automatically placed where the relevant context already lives.

How does it work?

When you set up a trigger, you select the Pod where its conversations should run. From then on, each time the trigger fires, the resulting conversation is created inside that Pod alongside your team's related work.

Concrete Use Cases

Here's how you could use it:

• Recurring reporting: A daily trigger that generates a summary now posts its conversation straight into your team's reporting Pod, where everyone already looks for updates.
• Automated monitoring: A trigger watching for specific events creates its conversations inside the dedicated project Pod, so the whole team can review and act on them in context.

Benefits for you

Less time spent hunting for automated runs, better visibility for your team, and cleaner organization of all triggered work — each conversation lands exactly where it belongs, with the right people able to access it.

How to access it?

This is available to all users. Simply open your trigger configuration and select the Pod where you'd like the triggered conversations to run.

Triggers Now Run Inside Pods

🎯 What is it?

When a trigger fires, the conversation it creates now lives directly inside the Pod you've selected. This means every automated run lands in the right workspace context from the start, rather than sitting outside your team's organized spaces.

💡 Why is it useful?

Automated work can quickly become hard to track when it's scattered outside your team's normal environment. By routing triggered conversations into a Pod, everything stays organized, easy to find, and shareable with the right people — automatically placed where the relevant context already lives.

⚙️ How does it work?

When you set up a trigger, you select the Pod where its conversations should run. From then on, each time the trigger fires, the resulting conversation is created inside that Pod alongside your team's related work.

✨ Concrete Use Cases

Here's how you could use it:

• Recurring reporting: A daily trigger that generates a summary now posts its conversation straight into your team's reporting Pod, where everyone already looks for updates.
• Automated monitoring: A trigger watching for specific events creates its conversations inside the dedicated project Pod, so the whole team can review and act on them in context.

📈 Benefits for you

Less time spent hunting for automated runs, better visibility for your team, and cleaner organization of all triggered work — each conversation lands exactly where it belongs, with the right people able to access it.

🚀 How to access it?

This is available to all users. Simply open your trigger configuration and select the Pod where you'd like the triggered conversations to run.

🎯 What is it?

We've added Anthropic's latest model, Claude Sonnet 5, to Dust. The claude-sonnet global agent now runs on Sonnet 5 automatically, and you can also select Sonnet 5 directly when building your own agents.

💡 Why is it useful?

You get access to the newest generation of the Sonnet model — with its improvements in reasoning, quality, and overall performance — without having to change any of your existing setups. The upgrade happens seamlessly behind the global agent.

⚙ How does it work?

The claude-sonnet global agent has been updated to point to Sonnet 5, so it's already using the new model. In the agent builder, Sonnet 5 now appears as an option in the model picker, ready to be assigned to any agent you create or edit.

✨ Concrete Use Cases

Here's how you could use it:

• Upgrade an existing agent: Edit one of your custom agents and switch its model to Sonnet 5 to benefit from the latest improvements.
• Quick everyday tasks: Simply use the claude-sonnet global agent — it already runs on Sonnet 5, no configuration needed.

📈 Benefits for you

Access to a more capable model for your agents, improved output quality, and a smooth transition since your current configurations keep working as-is.

🚀 How to access it?

If your workspace is on a paid plan, Sonnet 5 is already available — no feature flag required. Use it through the claude-sonnet global agent, or select it in the model picker within the agent builder.

⚠ Please note

Sonnet 5 is not yet available for workspaces with EU data residency. We'll keep you informed as soon as that changes.

What is it?

The Computer is a new capability that gives your agents access to an isolated, secure environment where they can work directly with files, run calculations, execute code and handle larger multi-step workflows. Optionally, you can authorize your Computer environment to make network requests to specific destinations.

Why is it useful?

Until now, agents were limited when it came to tasks that required actually processing files, running code, or chaining together more complex operations. The Computer removes those limits by giving agents a safe, sandboxed space to get hands-on work done, so they can support richer and more demanding tasks end to end.

How does it work?

Agents operate inside an isolated environment where they can safely manipulate files, perform computations, run code, and connect to external resources. Because the environment is sandboxed, this work happens securely without affecting your other systems. This also unlocks support for most external agent skills.

Concrete Use Cases

Here's how you could use it:

• File processing: Have an agent extract text from a PDF, clean up a spreadsheet, or convert and reorganize a set of documents.
• Calculations & data work: Ask an agent to run numerical analysis or process a dataset and return the results.
• Code execution: Let an agent write and run code to automate a task or test a quick script.
• Larger workflows: Combine several steps, such as fetching data via a network request, transforming it, and producing a final output, in a single agent run.

Benefits for you

More powerful agents and more efficient workflows. Tasks that previously required manual effort or couldn't be handled at all can now be completed directly by your agents, with broader support for external agent skills.

How to access it?

The Computer is available to everyone, with one exception: Enterprise workspaces that were not part of the beta. If that's your case, you can opt in at any time, and every workspace will automatically receive the feature within the next 2 weeks. You can learn more in the documentation.

What is it?

We've added Anthropic's latest model, Claude Sonnet 5, to Dust. The claude-sonnet global agent now runs on Sonnet 5 automatically, and you can also select Sonnet 5 directly when building your own agents.

Why is it useful?

You get access to the newest generation of the Sonnet model — with its improvements in reasoning, quality, and overall performance — without having to change any of your existing setups. The upgrade happens seamlessly behind the global agent.

How does it work?

The claude-sonnet global agent has been updated to point to Sonnet 5, so it's already using the new model. In the agent builder, Sonnet 5 now appears as an option in the model picker, ready to be assigned to any agent you create or edit.

Concrete Use Cases

Here's how you could use it:

• Upgrade an existing agent: Edit one of your custom agents and switch its model to Sonnet 5 to benefit from the latest improvements.
• Quick everyday tasks: Simply use the claude-sonnet global agent — it already runs on Sonnet 5, no configuration needed.

Benefits for you

Access to a more capable model for your agents, improved output quality, and a smooth transition since your current configurations keep working as-is.

How to access it?

If your workspace is on a paid plan, Sonnet 5 is already available — no feature flag required. Use it through the claude-sonnet global agent, or select it in the model picker within the agent builder.

Please note

Sonnet 5 is not yet available for workspaces with EU data residency. We'll keep you informed as soon as that changes.

What is it?

The "Ask user question" capability is now built into every agent by default. This means any agent can pause mid-task to ask you a clarifying question when it needs more information, instead of guessing or proceeding with incomplete instructions. Previously, this had to be manually enabled on a per-agent basis.

Why is it useful?

Until now, this capability was opt-in, which made it hard to discover and easy to forget when building a custom agent. By making it a default for all agents, you get more reliable, conversational interactions out of the box — agents can check in with you before heading in the wrong direction.

How does it work?

When an agent encounters ambiguity or needs a decision from you, it can now ask you directly during a task. You answer, and the agent continues with the right context. No setup or configuration is needed.

Concrete Use Cases

Here's how you could use it:

• Ambiguous requests: You ask an agent to "draft the email to the client" — instead of assuming which client, it asks you to confirm before writing.
• Missing details: An agent working on a report notices a required figure is missing and asks you for it, rather than leaving a gap or inventing a value.

Benefits for you

More accurate results, fewer back-and-forth corrections, and a smoother collaboration with your agents — especially for custom agents where this previously had to be configured manually.

How to access it?

There's nothing to set up. This is now active by default for all agents across your workspace, including any custom agents you've already built.

The Computer: Let Your Agents Work Safely with Files, Code, and More

🎯 What is it?

The Computer is a new capability that gives your agents access to an isolated, secure environment where they can work directly with files, run calculations, execute code and handle larger multi-step workflows. Optionally, you can authorize your Computer environment to make network requests to specific destinations.

💡 Why is it useful?

Until now, agents were limited when it came to tasks that required actually processing files, running code, or chaining together more complex operations. The Computer removes those limits by giving agents a safe, sandboxed space to get hands-on work done, so they can support richer and more demanding tasks end to end.

⚙️ How does it work?

Agents operate inside an isolated environment where they can safely manipulate files, perform computations, run code, and connect to external resources. Because the environment is sandboxed, this work happens securely without affecting your other systems. This also unlocks support for most external agent skills.

✨ Concrete Use Cases

Here's how you could use it:

• File processing: Have an agent extract text from a PDF, clean up a spreadsheet, or convert and reorganize a set of documents.
• Calculations & data work: Ask an agent to run numerical analysis or process a dataset and return the results.
• Code execution: Let an agent write and run code to automate a task or test a quick script.
• Larger workflows: Combine several steps, such as fetching data via a network request, transforming it, and producing a final output, in a single agent run.

📈 Benefits for you

More powerful agents and more efficient workflows. Tasks that previously required manual effort or couldn't be handled at all can now be completed directly by your agents, with broader support for external agent skills.

🚀 How to access it?

The Computer is available to everyone, with one exception: Enterprise workspaces that were not part of the beta. If that's your case, you can opt in at any time, and every workspace will automatically receive the feature within the next 2 weeks. You can learn more in the documentation.

What is it?

Pod editors can now manage an AGENTS.md file directly from their Pod settings. Once set, Dust automatically adds its content to the system prompt of every conversation that happens within that Pod, so your guidelines and workflows are always in effect.

Why is it useful?

Previously, keeping behavior consistent across multiple conversations often meant copy-pasting the same instructions over and over. This feature lets you define your Pod-level rules and workflows once, in a single place, and have them applied everywhere automatically—ensuring consistency across every run without the manual overhead.

How does it work?

Editors update the AGENTS.md file from the Pod settings, and Dust injects it into the system prompt for all conversations in that Pod. Oversized files are now handled more safely, so large instruction sets won't cause issues.

Concrete Use Cases

Here's how you could use it:

• Consistent tone and formatting: Define how agents should respond (language, style, structure) so every conversation in the Pod follows the same standards.
• Shared workflows and guardrails: Document the steps, do's and don'ts, or compliance rules your team must follow, and have them automatically enforced in each conversation.

Benefits for you

You save time by setting your rules once instead of repeating them, you reduce the risk of inconsistent behavior across conversations, and you gain a single, easy-to-maintain source of truth for how your Pod should operate.

How to access it?

This is available to all Pod users for Pod conversations—no feature flag required. Pod editors can simply open their Pod settings and edit the AGENTS.md file to get started.

Pods Now Apply Your AGENTS.md to Every Conversation

🎯 What is it?

Pod editors can now manage an AGENTS.md file directly from their Pod settings. Once set, Dust automatically adds its content to the system prompt of every conversation that happens within that Pod, so your guidelines and workflows are always in effect.

💡 Why is it useful?

Previously, keeping behavior consistent across multiple conversations often meant copy-pasting the same instructions over and over. This feature lets you define your Pod-level rules and workflows once, in a single place, and have them applied everywhere automatically—ensuring consistency across every run without the manual overhead.

⚙ How does it work?

Editors update the AGENTS.md file from the Pod settings, and Dust injects it into the system prompt for all conversations in that Pod. Oversized files are now handled more safely, so large instruction sets won't cause issues.

✨ Concrete Use Cases

Here's how you could use it:

• Consistent tone and formatting: Define how agents should respond (language, style, structure) so every conversation in the Pod follows the same standards.
• Shared workflows and guardrails: Document the steps, do's and don'ts, or compliance rules your team must follow, and have them automatically enforced in each conversation.

📈 Benefits for you

You save time by setting your rules once instead of repeating them, you reduce the risk of inconsistent behavior across conversations, and you gain a single, easy-to-maintain source of truth for how your Pod should operate.

🚀 How to access it?

This is available to all Pod users for Pod conversations—no feature flag required. Pod editors can simply open their Pod settings and edit the AGENTS.md file to get started.

What is it?

GLM 5.2 is a new AI model developed by the lab Z.ai, now available within Dust. It performs at a frontier level on many agentic benchmarks, with results comparable to leading models like Opus 4.8 and GPT 5.5. The model is hosted in the United States by Fireworks.

Why is it useful?

Having access to a broad range of high-performing models means you're never locked into a single provider. GLM 5.2 expands your options with a model that delivers top-tier performance on agentic tasks, giving you more flexibility to choose the right model for your specific needs and use cases.

How does it work?

GLM 5.2 is available as a model option when building or configuring your agents. You simply select it from the list of available models, just as you would with any other supported model. Although developed by Z.ai, it is hosted in the US by Fireworks.

Concrete Use Cases

Here's how you could use it:

• Agentic workflows: Power agents that handle multi-step tasks and tool use, where GLM 5.2's strong agentic benchmark performance shines.
• Model comparison: Test GLM 5.2 against your current models on your own use cases to see which delivers the best results for your team.

Benefits for you

More choice, more flexibility. You gain access to an additional frontier-level model, allowing you to optimize your agents for performance while diversifying beyond a single model provider.

How to access it?

GLM 5.2 is now available in the model selection menu when creating or editing an agent. Simply choose it as your model to start testing it on your workflows.

Adomik Integration: Ad Revenue Analytics & Knowledge, Directly in Dust

What is it?

Adomik is now available as a default integration in Dust, connected securely via OAuth. Once enabled, your builders can give agents read-only access to Adomik's tools, allowing them to analyze your programmatic advertising and monetization data, as well as search and retrieve pages from your Adomik knowledge base, all without leaving Dust.

Why is it useful?

If your teams work with programmatic advertising, investigating revenue movements or pricing shifts often means jumping between tools and manually pulling reports. This integration brings that data and knowledge directly into your agents, so you can run monetization investigations and draft analyses in one place.

How does it work?

Adomik connects as a remote integration through a secure OAuth login. The tools are strictly read-only, meaning agents can analyze and retrieve your data but cannot modify anything on the Adomik side.

Concrete Use Cases

Here's how you could use it:

• Revenue investigation: Ask an agent to identify and explain sudden CPC/CPM shifts or revenue movements across your programmatic inventory.
• Report drafting: Have an agent pull the relevant monetization figures and draft a first version of a performance report for your team.
• Knowledge lookup: Let an agent search your Adomik knowledge base to quickly answer process or methodology questions.

Benefits for you

You save time by consolidating ad data analysis and knowledge retrieval into your existing Dust agents, reduce manual back-and-forth between platforms, and give your teams faster, data-backed answers on monetization questions.

How to access it?

An admin first connects Adomik from Spaces → Tools. Once connected, your builders can add the Adomik tools to any relevant agent.

What is it?

The Lemlist MCP Server is now available across all Dust workspaces. It connects your Dust agents directly to Lemlist, giving them access to your leads, verified contacts, and multichannel outreach sequences. In short, your agents can now help you manage your outbound workflow without you having to switch tools.

Why is it useful?

Outbound prospecting often means juggling several tools and repeating the same manual steps: pulling lead lists, checking contact details, updating sequences. By bringing Lemlist into Dust, you can hand those tasks to an agent and keep your focus on strategy and conversations that matter.

How does it work?

Once enabled, an agent can query and act on your Lemlist data through the connection: retrieving leads, accessing verified contact information, and working with your multichannel sequences. You simply ask the agent in natural language, and it interacts with Lemlist on your behalf.

Concrete Use Cases

Here's how you could use it:

• Lead review: Ask an agent to pull a list of leads from a given Lemlist campaign and summarize who's worth prioritizing.
• Contact enrichment: Have an agent retrieve verified contact details for a set of prospects before you reach out.
• Sequence management: Ask an agent to check the status of your multichannel sequences or help you organize prospects across channels.

Benefits for you

Less tool-switching, faster access to your outbound data, and the ability to delegate repetitive prospecting tasks to an agent. The result is a smoother outbound workflow and more time for high-value selling.

How to access it?

The Lemlist MCP Server is available in all Dust workspaces. You can add it as a tool when building or editing an agent, then connect your Lemlist account to start using it. If you're a Lemlist user, it's ready for you to set up today.