Skip to main content

Changelog

Follow new updates and improvements to Swiftask.

Fewer back-and-forth steps, more time for what matters most

Hello,

Two new features at the start of this week, both centered around a single idea: fewer detours to get what you need.

💻 Your agent can now use your computer

New — in beta, available in the desktop app for Mac, Windows, and Linux.

Some data is only accessible from your workstation: an internal database, a business tool, or a company API. Reaching them often requires being connected to the VPN.

Your agent can now go there for you. You ask your question in the chat, it fetches the info, and it answers you—without opening three different tools and without bothering the tech team.

How to get started in 3 steps

1. Install the Swiftask app on your computer. This feature is only available on the desktop version—it needs to be installed on your workstation to access it. Download it here: swiftask.ai/fr-fr/ai-app-multi-device (Mac, Windows, Linux). Installation takes less than a minute.

2. Open a conversation, then click the computer icon in the toolbar below the message box. The Computer Access panel will open and display your workstation under the Local Devices section.

3. Click “Save and use in this chat.” Done: your agent can now work from this workstation, for this conversation only.

Nothing happens without you. Nothing is enabled by default. You can cut off access whenever you want, and your agent asks for your permission before any major action.

To try it out: Once access is enabled, ask for information stored in an internal company tool.

📌 Pin your important conversations

The chat you reopen ten times a day will no longer slip down the list.
Pin it: it moves up to a "Pinned" section at the very top of your sidebar and stays there. Handy for your current project, a client follow-up, or the conversation you pick up every morning.

How to do it: Click the three dots to the right of a chat, then click Pin. Use the same steps to unpin it.

To try it out: Pin your most active conversation, then open several new chats. It will still be right at the top.

3 New Updates on Swiftask: Visual Responses, Video Generation by Gemini Omni, and Three Cost-Effective Models

Hello 👋

This week, your agents get a new way to display information, learn how to create videos, and three new models join the lineup to help lower your bill.


🖼️ Interactive Interface: Responses you look at, not just read

Your agents can now respond visually right inside the conversation, rather than using text alone.

Ask for the weather in Paris, and you’ll get an interactive weather widget—14°C, sunny, hour-by-hour forecast—with three tabs: Today / 5 Days / Rain. You click, and the display updates without having to ask another question.

The same applies to your business use cases: monthly figures with Month / Quarter / Year tabs, a scrollable comparison of three competitors, or a support agent displaying pricing and hours cleanly instead of a dense wall of text. Information is faster to grasp, easier to digest, and far more impactful.

You decide when to use it. In the skill settings, a Description field lets you specify when the agent should switch to visual mode: comparing multiple options as cards, gathering preferences through a short form, displaying browsable metrics… You stay in control of the timing.

Where to enable it: Agents → Your agent → Agent Instructions → Skills → Interactive Interface.

🎬 Gemini Omni: Create videos just by chatting

Your agents can now generate videos using Gemini Omni (the successor to Veo): 3- to 15-second clips with simultaneous audio track generation (ambient sound, voiceovers, and synchronized dialogue).

The real standout feature is natural-language video editing: no need to regenerate the entire video from scratch. Provide a reference image (a product photo, for example), request a camera movement, then prompt "change the background to wood," and your product remains intact. It’s also great for creating a vertical teaser from your latest post or an illustrative clip for a training module.

Save your credits by generating drafts in 360p or 720p, and only switch to 1080p/4K for the final version. You can also extend a video up to 30 seconds.

Where to find it: Agents → Configure agent → Skills → Video Generation (Gemini Omni).


🧠 Three Models Hosted by Mistral, Built for Agents

Three new models are joining your library, all hosted by Mistral (with European routing available) and built for agentic workflows, each featuring a 1M context window.

  1. GLM 5.3 Flash (Z.ai): The best value for money. Multimodal (text, image, video), it rivals top-tier models on agentic tasks at roughly one-tenth the cost. Ideal for high-volume agents: customer support, email sorting, and monitoring.

  2. GLM 5.3 (Z.ai): The heavy-duty version, built for deep reasoning: RFPs, technical documentation, and multi-step projects.

  3. DeepSeek V4.1 Flash: Fast and versatile, supporting text and images. Great for internal assistants and real-time interactions.

Use case idea: Your agent that qualifies 300 leads a month doesn't need a flagship model. Switch it to GLM 5.3 Flash, and save the high-end model for the agent writing your sales proposals. The right model in the right place.

Where to enable them: Chat → Switch AI, or Agents → Your agent → LLM. Available on all plans at no extra cost.

What’s new in November? More control, greater clarity

Hello,

Four new updates to kick off the month.

1. An indicator tells you when to open a new conversation

A conversation that gets too long eventually loses accuracy: the AI forgets an instruction, mixes up details, or hallucinates. A percentage is now displayed at the bottom of the conversation — when it goes up, open a new one and start fresh.

The rule of thumb: one conversation per topic.

2. Redesigned Artifacts, and true control over published agents

The Artifacts page is clearer to browse as presentations and web pages accumulate.

On the governance side, sharing is managed from Roles, and publishing an agent becomes a permission in its own right. Putting an agent online makes it accessible to people without a Swiftask account, with your knowledge base behind it and your credits running: this action is now restricted only to the people you designate.

The new External Public Agents page lists all your published agents — handy for spotting one that was forgotten.

Where to find it: Settings → Identity → Role → Permissions · Settings → Workspace Design → External Public Agents

3. Your workspace has its own login page

Your web address, your logo, and only the methods you choose (email, Google, Microsoft, or your SSO). No more hesitation when logging in, and a single link to send to newcomers.

Where to find it: Workspace Authentication Settings

4. AI consumption, now visible by group

You can finally see which teams use AI the most, and which ones haven't started yet. Check it once a month: it's often where you discover that a single use case accounts for half the budget.

Where to find it: Dashboard → Usage by Group

Head to your workspace to test all this out. More updates coming soon 😉

Earlier updates