Changelog

You can now choose between three frontier models based on your specific needs and budget. All models are available immediately via the Chat Model Selector and Agent Configuration, across all plans.

Kimi 2.7 Code (Moonshot AI)

High performance, controlled budget.

Moonshot has created a model that delivers exceptional results while using 30% less computing power than its predecessor. Perfect for teams deploying high volumes of agents and workflows.

• 0.6 credit/word: the most economical choice in this selection

• Excellent performance for complex tasks and multi-step workflows

• Ideal for teams looking to scale AI agents while keeping costs low

GLM-5.2 (Z.ai)

Deep understanding, massive context, fair pricing.

Z.ai has built a model capable of reading entire projects in one go. It is ideal if you need to analyze long documents, massive codebases, or coordinate complex, multi-step workflows.

• 0.9 credit/word: the sweet spot between cost and capacity

• Capable of absorbing the equivalent of a small library in a single request

• Great for long-document analysis and long-running autonomous workflows

Claude Opus 4.8 (Anthropic)

Premium reliability for critical tasks.

Anthropic has delivered its most reliable model yet for high-stakes decision-making. It is 4x better at detecting its own errors than the previous generation and "thinks" before acting on complex problems.

• 3.48 credits/word: our premium model

• Recommended for enterprise workflows where precision is non-negotiable

• Ideal for tasks requiring deep reasoning, high accuracy, and professional-grade reliability

How to choose?

• High-volume tasks on a budget? Start with Kimi 2.7 Code to optimize costs.

• Complex projects or massive documents? Go with GLM-5.2 for its massive context capacity.

• Critical enterprise decisions? Invest in Claude Opus 4.8 for maximum reliability.

You can use all three simultaneously in your workspace, simply pick the best model for each specific agent.

Improvements

• Real-time model switching: Update your models in Agent Configuration instantly without needing to refresh the page.

• Enhanced model cards: The Chat Model Selector now displays costs and key capabilities at a glance, making it easier to choose the right tool for the job.

Get started now: Open your Chat Model Selector or create a new agent to test your first frontier model.

Admins can now assign temporary credit to individual workspace members, handling seasonal spikes and urgent requests without changing permanent credit limits. Set an expiration date, and the temporary allowance automatically resets when it expires.

Temporary credit limits for individual members

Give any team member an immediate credit boost without modifying their permanent monthly limit. Perfect for handling unexpected demand, seasonal projects, or helping someone finish an important task. The temporary credit sits on top of their existing limit and expires on a date you control.

• Set credit amount: Choose how much temporary credit to grant, up to your workspace's remaining balance

• Customize expiration: Default expires at the end of the month, but you can set any date you want

• One-time override: Temporary credit replaces a member's monthly limit only until it expires; their permanent limit stays unchanged

• Apply to multiple users: Use the "Apply to all users" toggle to grant the same temporary credit to your entire team at once

• No extra cost: Temporary credit comes from your workspace's balance; no additional charges

Where to find it: Settings → Workspace → Members, click the edit icon (pencil) next to a member's Credit Limit, then toggle "Temporary credit limit" on.

How to test it

1. Go to Settings → Workspace → Members

2. Click the edit icon next to any member's credit limit

3. Toggle on "Temporary credit limit"

4. Enter a credit amount (up to your remaining workspace balance)

5. Set the expiration date (default is the last day of this month)

6. Click Validate

7. Verify the member's credit limit now shows the temporary amount

8. Confirm the limit resets on the expiration date

Available on: All plans (Pro, Team Starter, Team Growth, Tailored)

You can now enable or disable all skills, LLMs, and agents in a single click when configuring a role. No more managing permissions one by one.

Where to find it: Settings → Workspace → Roles → [Select a role] → Skills / LLM Models / Agents

What's included:

• "Activate all" button to enable all available skills, LLMs, and agents at once.

• "Deactivate all" button to start from scratch.

• Real-time counter: "X items enabled out of Y available."

• Full flexibility: You can still edit individual permissions after a bulk action.

• Auto-save: All changes are saved automatically.

Available on: All plans (Pro, Team Starter, Team Growth, Tailied/Enterprise)

How to test

1. Go to Settings → Workspace → Roles.

2. Select or create a custom role.

3. Click "Activate all" in the Skills section — all toggles turn on.

4. Check the counter: "16 skills enabled out of 16 available."

5. Click "Deactivate all" — all toggles turn off.

6. Manually enable a few skills to confirm individual toggles still work.

7. Assign a user to this role and verify their access in the agent builder.

Improvements

Real-time counter — Instantly see how many skills, models, or agents are authorized for the role.

Faster bulk actions — Role configuration is now 5x faster.

Full flexibility — You can still fine-tune individual permissions after a bulk action.

Note: No changes for existing roles. All current roles keep their current permissions.

We're excited to announce a game-changing addition to the agent builder: you can now create fully configured AI agents by describing what you want in plain language. Instead of manually building agents from scratch, simply tell Swiftask what your agent should do, and our AI agent builder generates a working agent in seconds, ready to customize and deploy.

Create with AI: Describe your agent, get a working agent

Create custom AI agents faster than ever by typing a simple description. Swiftask analyzes your request and automatically generates an agent with everything pre-configured: name, description, system instructions, icon, recommended LLM models, and suggested skills.

How it works:

• Go to Agents → Create agent

• Select the "Create with AI" option in the creation modal

• Describe what your agent should do (e.g., "Create a marketing expert agent that analyzes campaigns and suggests improvements")

• Swiftask instantly generates a fully configured agent with:

  • Auto-generated name, description, and icon

  • System instructions tailored to your use case

  • 5 recommended LLM models to choose from

  • Suggested skills and capabilities

  • Complete agent configuration ready to use

Refine in real-time with the agent builder:

Once your agent is created, you can chat directly with the AI agent builder ("Le Maestro") to make changes:

• Click "Use this LLM" on any recommended model to integrate it instantly

• Ask the builder to suggests what skills to integrate in the agent

• Request configuration changes without leaving the creation interface

• The builder applies changes on the fly

Then customize manually:

After AI generation, you can manually edit every aspect of your agent, just like building from scratch:

• Agent name, description, and icon

• System instructions and objectives

• Skills and knowledge base

• AI model selection

• Any other configuration

Test immediately:

Once created, test your agent right away using the built-in test panel. See how it responds to prompts and refine before deployment.

Availability: Available on all plans (Pro, Team Starter, Team Growth, Custom)

How to test this feature

1. Navigate to Agents in the left sidebar

2. Click the Create agent button

3. Choose "Create with AI" from the three creation options

4. Write a prompt describing your desired agent (e.g., "Create a customer support agent that handles billing questions and escalates complex issues")

5. Review the generated agent configuration

6. Click on any LLM suggestion and select "Use this LLM" to apply it instantly

7. Ask the AI builder to suggest skills or configuration as needed (e.g., "What are the best skills for this agent")

8. Click Test & Preview to interact with your newly created agent

9. Make manual edits in Manual setup if needed

10. Deploy and start using your agent

What this means for you

• Faster agent creation: Build agents in seconds instead of hours of manual configuration

• Less guesswork: The AI recommends the best LLM and skills based on your use case

• Guided customization: Discuss your needs with the AI builder and see changes applied instantly

• Full control: Manually edit anything after creation, or let AI do the heavy lifting

• No technical skills required: Describe your agent in plain language; no code needed

The "Create with AI" option sits alongside existing methods (From Scratch, From Template), giving you three flexible paths to agent creation depending on your needs.

Ready to build your first AI agent with AI? Open Swiftask, navigate to Agents, and try creating an agent with AI today. Your next specialized AI assistant is just a description away.

We've redesigned the Chat AI browsing and filtering interface to make discovering and selecting the right model faster and more intuitive. Whether you're looking for speed, advanced reasoning, or data sovereignty, the new interface gets you there with clarity and confidence.

Clearer filtering interface

Filter AI models by what matters most: provider, privacy, and region. Three dedicated dropdowns at the top of the Chat section let you narrow down your choices instantly, without guessing or scrolling endlessly.

• Provider filter: Find models by vendor (OpenAI, Google, Anthropic, Scaleway, and more)

• Privacy filter: Select models based on data handling and compliance needs

• Region filter: Choose models hosted in your preferred data center or region

• All filters work together, so you can combine them for precise results

Find it in: Chat → Browse AI models

Improved visual hierarchy and readability

Each AI model now appears as a clean, scannable card in a 2-column grid. You see exactly what matters at a glance:

• Model icon and name: Instantly recognizable by provider

• Brief description: What the model excels at (truncated for space, full text on hover)

• Cost information: Transparent pricing per token or word

• Provider and region badges: Security shields and flag icons show data handling and location at a glance

• Clear visual separation: Sections for Chat models, Agent templates, and Workspace pinned agents are visually distinct

Faster access to key information

No more digging. Costs, descriptions, and provider details appear directly on each model card, so you can compare options without clicking into details.

• Model pricing is always visible

• Provider identity is clear via icons and labels

• Privacy and region information appears as badges

• Descriptions are concise and helpful

Optimized navigation between sections

The horizontal tab system makes it effortless to move between Chat models, Agent templates, Integrated tools, and other sections. Each tab is clearly labeled, and the active tab is visually highlighted.

• Tabs: Recent, Workspace pinned agents, Integrated tools, My agents, Chat, Agent templates, Web, Image, Audio

• "View more" button expands lists without leaving the interface

• Back and Close buttons provide clear exit paths

Enhanced visual feedback

Icons, colors, and badges now guide you through the interface:

• Security badges: Shield icons indicate privacy and compliance features

• Region indicators: Flag icons show where data is hosted

• Color coding: Visual distinction between model types and sections

• Interactive states: Buttons and filters respond clearly to your actions

How to test it

1. Open Chat from the left sidebar

2. Look for the browsing panel showing available AI on the “Change AI” button

1. Try the filters: Click Provider: All and select a specific vendor (e.g., "OpenAI")

2. Click Privacy: All to see privacy options

3. Click Region: All to filter by data location

4. Explore the model cards and notice how cost, description, and provider information is organized

5. Switch between tabs (Chat, Agent templates, etc.) using the horizontal navigation

6. Click on a model card to select it and start chatting

Availability: All Swiftask plans, Pro, Team Starter, Team Growth, and Custom. No extra cost.

Ready to find your ideal AI model? Open Chat now and explore the new browsing interface. It's faster, clearer, and more beautiful.

Take control of enterprise security, one login at a time.

We're excited to introduce a new security setting that lets workspace admins enforce SAML authentication across their organization. If your team relies on corporate identity providers like Azure AD or Okta, you can now require all new users to sign in through your enterprise SSO, eliminating password-based access for those accounts.

Enforce SAML authentication

Admins can now toggle on "Enforce SAML Authentication" to mandate that all new users with your configured email domain sign in exclusively through your identity provider. This prevents password-based logins for those accounts and keeps your enterprise access tightly controlled.

How it works:

• When enabled, new users with email addresses matching your configured domain must authenticate via SAML SSO

• Existing users can continue using their current password (no forced migration)

• All users, new or existing, must authenticate to access Swiftask (no session persistence without login)

• Works seamlessly with your existing SAML configuration

Where to find it:

Navigate to Workspace settings → SSO/SAML and toggle "Enforce SAML Authentication" on. You'll see a clear red banner confirming the setting is active.

Plan availability:

Available on Team Growth, Team Starter, and Custom plans (same as SSO/SAML configuration)

How to test it

1. Enable the setting:

   • Go to Workspace settings → SSO/SAML

   • Toggle "Enforce SAML Authentication" on

   • Confirm your email domain is set (e.g., company.com)

   • Click "Save Configuration"

2. Test with a new user:

   • Invite a team member with an email matching your configured domain

   • Have them attempt to log in with their email and password

   • They should see a SAML SSO login option instead

   • Confirm they can authenticate through your identity provider

3. Verify existing users:

   • Log in with an existing account that has a password

   • Confirm access still works (backward compatible)

   • Note that they'll still need to authenticate each session

4. Check the Auto Join setting (optional):

   • If "Auto Join Users" is enabled, new SSO users are automatically added to your workspace

   • If disabled, admins must manually invite users

Why this matters

Enforcing SAML authentication gives your organization:

• Stronger security: Corporate identity provider controls authentication, not individual passwords

• Centralized access control: Manage user access through your identity provider

• Compliance ready: Meets enterprise security requirements for regulated industries

• Backward compatible: Existing users can keep using their passwords; new users follow the new rule

• No disruption: Roll out gradually without forcing immediate changes

Ready to secure your workspace?

Navigate to Workspace settings → SSO/SAML and enable "Enforce SAML Authentication" today.

We've redesigned the "Coming up" events list to help you manage your calendar and meetings in fewer clicks. Click any event to instantly access recording options, meeting links, and assistant selection, everything you need to join or record your next meeting is now just one tap away.

Improved "Coming up" events display

The events list is now clearer and easier to scan. Each event shows the date, time, title, and participant count at a glance, making it simple to find what you're looking for without opening individual calendar entries.

• Cleaner layout: Date, time, and event title are prominently displayed

• Quick visual scanning: Participant indicators show who's attending

• Organized by date: Events grouped chronologically for easy navigation

• Find it in → Calendar → Coming up (sidebar)

One-click meeting actions

Click any event to open a modal showing all your meeting options. No more hunting for meeting links or switching between apps.

What you can do from the event modal:

• View meeting details: See the full event title, date, time, and participant list

• Copy meeting link: Grab the Google Meet, Teams, or Zoom link with one click

• Join with an agent: Start the meeting with your AI assistant already present (available on all plans)

• Record locally: Capture the meeting directly from your Swiftask desktop app with full transcription and AI analysis

• Select a meeting assistant: Choose which AI assistant joins the meeting, or set a default for future meetings

• Visual feedback: See loading indicators and confirmation messages as actions complete

Supported platforms: Google Meet, Microsoft Teams, Zoom

Find it in → Calendar → Coming up → Click any event

Quick access to meeting recording

Whether you want to record locally on your desktop or have the AI assistant join remotely, both options are now visible and accessible from the same place.

• Record locally: Use the Swiftask desktop app to capture audio, transcription, and AI-powered insights

• Join with agent: The AI assistant joins your meeting as a participant, visible to all attendees, and handles transcription and analysis automatically

• Billing transparency: See token costs upfront (e.g., "Meeting transcription is billed at 4.5K tokens per minute")

• Assistant selection: Pick from available AI assistants or set one as your default for all future meetings

• Available on all plans: No additional cost; meeting management is included in your Swiftask subscription

How to test it

1. Navigate to your calendar: Open Swiftask and go to Meeting → Coming up

2. Click any upcoming event: A modal will appear with all meeting details and action buttons

3. Try the actions:

   • Click the link icon to copy the meeting URL

   • Click "Join with agent" to start the meeting with AI assistance

   • Click "Record locally" to launch the desktop recording interface

   • Use the dropdown to select or set a default AI assistant

4. Watch for feedback: Notice the loading indicators and success messages as actions complete

5. Test across platforms: Repeat with events from Google Meet, Teams, and Zoom to verify consistency

Improvements

Better event readability: The "Coming up" list now uses clearer typography and spacing, making it easier to spot upcoming meetings at a glance, especially when you have multiple events on the same day.

Faster meeting setup: No more copy-pasting links or navigating to separate settings. Everything you need to join or record a meeting is accessible directly from the event modal, cutting setup time from multiple clicks to one.

Transparent billing: Token costs for meeting transcription are now visible in the modal before you start recording, helping you understand exactly what you'll be charged.

Consistent assistant selection: The meeting assistant dropdown now includes a "Set as default" checkbox, so your preferred assistant is automatically selected for future meetings without manual intervention each time.

Ready to try it?

Update your Swiftask app now to see the new "Coming up" events interface. Click any event and experience faster, smarter meeting management.

Questions? Check out our meeting assistant guide or visit the help center.

We're excited to introduce voice mode beta in chat, making it easier than ever to interact with your agents naturally. No more typing, no transcription steps, just speak and get instant answers.

Voice mode: Natural conversation with your agents

You can now chat with any agent using your voice. Click the voice icon next to the microphone in the chat input, speak your question or request, and your agent responds immediately with a text answer. It's that simple, no need to review transcriptions, no manual send button, no extra steps. Perfect for quick questions, hands-free work, or when typing isn't convenient.

• Available everywhere: Works with all agents (including your custom agents) and all LLM models, across all Swiftask plans

• Supported languages: All languages supported by Swiftask meeting recording

• Zero setup for users: Voice mode is ready to use directly in Chat → just click and speak

• Configurable per agent: For your custom agents, enable or disable voice mode on an agent-by-agent basis via the Agent Voice Settings menu.

• Visual feedback: A "Speak now" indicator appears when voice mode is active, so you always know when Swiftask is listening

• Browser-friendly: No browser restrictions, use voice mode on desktop, tablet, or any device with microphone access

How to test voice mode

For Users:

1. Open any agent chat in Swiftask.

2. Locate the voice icon (red circle) next to the microphone icon in the chat toolbar.

3. Click the voice icon to activate the mode.

4. Speak your question or command clearly.

5. Watch the "Speak now" indicator while you talk.

6. Once you finish speaking, the agent responds immediately in text.

7. Continue the conversation via voice, or switch back to typing whenever you like.

For Admins (Custom Agent Configuration):

1. Access your custom agent.

2. Navigate to Agent Voice Settings.

3. Enable or disable voice mode according to your needs.

4. Save your changes.

Pro tip: Voice mode works great for brainstorming, quick lookups, and hands-free workflows. Try asking your agent for advice, summaries, or code help, all without typing a single word.

What's next for voice mode

This is the beta launch. We're working on voice responses from agents (so your agent can talk back to you), and we'll continue refining based on your feedback. Let us know how voice mode works for you!