Changelog

When an agent provides a very long response, you no longer need to manually scroll to read it in its entirety. A new arrow button allows you to jump directly to the end of the response, making it easy to copy the content or view the conclusion effortlessly.

Quick navigation arrow to the end of responses

When an agent generates a long response, the chat now displays a navigation arrow that allows you to jump directly to the end of the message. No more tedious scrolling, one simple click and you are at the very end of the response.

What's new:

• Arrow visible on long responses: An arrow button appears whenever a response exceeds the screen height.

• Instant access to the end: Click the arrow to jump directly to the last word of the response.

• Easier copying: Once at the end, easily copy the full content of the response.

• Available everywhere in the chat: This improvement works across all Swiftask chats.

Where to find it: Open a chat → Ask the agent a question → Wait for a long response → Click the arrow to go to the end.

How to test:

1. Open a chat in Swiftask.

2. Ask a question that generates a long response (e.g., "Write a 500-word article about...").

3. Observe the navigation arrow that appears.

4. Click on the arrow.

5. Verify that you have reached the end of the response.

6. Easily copy the content.

Available on: All Swiftask plans.

Improvements

Optimized chat scrolling behavior to improve accessibility to long responses and facilitate content copying.

Starting today, all your new agents come equipped with two essential skills: Python data analysis and web search. No more bare-bones agents, get ready for assistants that can process your files and explore the web without any additional configuration.

Two powerful skills, built in by default

Every new agent you create now includes:

• Python: Code + Upload + Visualization: Import data files (Excel, CSV, JSON), execute Python code to transform them, generate charts, and download results in one click. Perfect for data analysis, cleaning, and visualization.

• Web Search: Access real-time information, scrape web pages, and enrich your analysis with the latest data from the web.

Supported File Types

With the Python skill, your agents can process a wide range of formats:

• Spreadsheets: XLSX, XLS, CSV

• Data Formats: JSON, XML, Markdown

• Images: PNG, JPG, GIF

• Code: Python, JavaScript, and other programming files

Where to Find Them

Access these skills in Agents → [Agent Name] → Skills. They are ready to use, no setup required.

Plan Availability

Both skills are available on all plans: Pro, Team Starter, Team Growth, and Custom.

How to Test

1. Create a new agent (or open a recent one).

2. Open the Chat and start interacting with your agent.

3. Test Python:

   • Import an Excel file → ask the agent to analyze it and create a summary

   • Request a data visualization

4. Test Web Search:

   • Ask for the latest news on a topic

   • Ask a question that requires current information

5. Observe the results: Generated files will appear with direct download links.

Common Use Cases

• Data Analysis: Import sales reports, let the agent extract key figures, and generate a chart.

• Research & Synthesis: Ask the agent to search for the latest market trends and summarize them.

• Data Cleaning: Import a messy CSV list, and the agent will structure it and return a clean Excel file.

• Information Enrichment: Combine local data with web searches to create comprehensive reports.

Impact on Existing Agents

Existing agents are not affected by this update. These skills are automatically added to all new agents created as of June 8, 2026.

If you want to add these skills to an existing agent, simply go to Agents → [Agent] → Skills → Add a skill and select the one you need.

Ready to try? Create a new agent and start analyzing your data and exploring the web immediately. No more friction, no more limitations, just intelligent agents, ready to handle your data.

We've added two new configuration options to give you more control over how your shared agents appear and behave when embedded on your website or app. Customize the conversation experience to match your needs.

Hide the chat history sidebar

You can now toggle the conversation history panel on or off in your embedded agent widget. When disabled, the chat takes up the full width, giving more screen space to longer conversations.

• Find it in: Agent profile → Advanced options → "Hide chat history sidebar"

• Enabled by default: Off (history sidebar visible)

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

Control skill visibility in conversations

New granular control over how skill executions (tools, thinking steps, etc.) appear in the chat. Choose to show everything or hide specific skill details to reduce visual clutter.

• Find it in: Agent profile → Advanced options → "Skills display in chat"

• Options: "Show everything" (default) or custom visibility per skill

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

Why this matters

These improvements apply to shared agents embedded via iframe, perfect for teams deploying agents on public-facing websites or internal portals.

• Maximize screen space for conversation-heavy workflows

• Reduce cognitive load with cleaner interface options

• Match your brand's UX preferences

• No code changes needed : configure once, deploy everywhere

How to test it

1. Go to your agent profile

2. Scroll to Advanced options

3. Toggle "Hide chat history sidebar" to remove the session list from your embedded widget

4. Toggle "Skills display in chat" to control how tool executions and thinking steps appear

5. Preview the changes directly in your embedded widget

Available now on all plans.

We're expanding Swiftask's multi-LLM capabilities with five new models, giving you more options to optimize for speed, reasoning depth, and cost. Switch between models instantly: no vendor lock-in, full control.

New models available

You can now use five new language models across all agents and chat interfaces:

• Deepseek 4 Flash: Fast, efficient model for quick responses and real-time interactions

• Deepseek Pro: Advanced reasoning and complex task handling

• Gemini 3.5 Flash: High-speed responses optimized for productivity workflows

• Scaleway-hosted Gemma 4: On-premise alternative for data sovereignty and compliance

• Scaleway-hosted Qwen 3.6: Flexible, open-source model with European hosting

Available on all plans: Pro, Team Starter, Team Growth, and Custom. No extra cost.

Find them here:

• Chat interface → Model Selector dropdown

• Agents → Agent Configuration → Model Selection

• Workspace Settings → Advanced → Default Model (to set workspace-wide defaults)

Seamless model switching

Switch between any model at any time—mid-conversation, between tasks, or at the agent level. Each model excels at different tasks:

• Use Deepseek 4 Flash or Gemini 3.5 Flash for quick answers and real-time interactions

• Use Deepseek Pro for deep analysis, reasoning, and complex problem-solving

• Use Scaleway-hosted models (Gemma 4, Qwen 3.6) when you need data to stay within Europe or prefer open-source alternatives

No downtime, no reconfiguration needed.

Authentication for Scaleway-hosted models

To use Scaleway-hosted Gemma 4 and Qwen 3.6, you'll need to configure your Scaleway API credentials:

• Location: Workspace Settings → Integrations → Scaleway

• Setup: Provide your Scaleway API key and endpoint

• Note: Not specified; exact authentication flow and whether this is per-model or per-provider configuration. Contact support for setup assistance.

Full compatibility with existing workflows

All five new models work seamlessly with:

• Task automation (trigger agents, execute workflows)

• Chat interfaces (direct conversations, pinned agents)

• API integrations (connect to your tools and data)

• Skills and knowledge bases (all existing agent capabilities)

No breaking changes. Your existing agents and workflows continue to work as before.

How to test the new models

1. In Chat:

   • Open the Chat interface

   • Click the Model Selector dropdown (top-right or near the input field)

   • Select one of the new models (Deepseek 4 Flash, Deepseek Pro, Gemini 3.5 Flash, Scaleway Gemma 4, Scaleway Qwen 3.6)

   • Send a test message and observe response quality and speed

2. In Agents:

   • Open an agent or create a new one

   • Navigate to Agent Configuration → Model Selection

   • Choose a new model from the dropdown

   • Test the agent with sample inputs that match your use case (e.g., fast queries for Flash models, complex reasoning for Pro models)

3. For Scaleway-hosted models:

   • First, configure your Scaleway credentials in Workspace Settings → Integrations → Scaleway

   • Then select Scaleway Gemma 4 or Qwen 3.6 in the model picker

   • Test with data-sensitive workflows to confirm data residency in Europe

4. Performance testing:

   • Compare response times and quality across models for your typical use cases

   • Monitor token usage if relevant to your plan

   • Switch models mid-task to find the best fit for each workflow

What's coming next

We're committed to expanding Swiftask's model library. More models and providers will be added regularly as new options become available.

Any questions? Contact support or visit the Swiftask help center for detailed setup guides for Scaleway-hosted models.

We're excited to introduce a new analytics dashboard for AI agents. You can now track usage, performance, and errors for each agent over the last 30 days, giving you the visibility you need to optimize deployments and understand where your credits are going.

Agent analytics dashboard

Every agent now has its own analytics hub, showing you exactly how it's performing. Four key metrics give you an instant snapshot: total credits consumed, total runs executed, completion rate, and active users. Scroll down for deeper visualizations that reveal trends, patterns, and performance over time.

Find it here: Agent → Agent modification space → Left sidebar → Analytics (under Deployment)

What you'll see:

• Total credits : Credits consumed by this agent in the period

• Total runs : Number of times the agent was executed

• Completion rate : Percentage of successful runs, with failure count

• Active users : Number of unique users who triggered the agent

• Run volume by channel : Bar chart breaking down runs by Swiftask vs. other channels

• Active users trend : Daily active user count over the 30-day window

• Credits over time : Line chart showing credit consumption patterns and spikes

• Avg credits per run : Efficiency metric showing average cost per execution

• Run status over time : Stacked bar chart comparing successful vs. failed runs by day

All data covers the last 30 days and updates in real-time as your agent runs.

Available on: All plans

How to test it

1. Go to any AI agent in your workspace

2. Open the agent modification space (click to edit)

3. In the left sidebar, scroll to the Deployment section

4. Click Analytics at the bottom

5. Verify the 4 metric cards display with data from your agent

6. Scroll down to confirm all 5 visualizations are rendering correctly

7. Run your agent a few times and check that metrics update in real-time

Your meetings now happen everywhere you do. Whether you're in Google Meet, Teams, or checking transcripts from the web, Swiftask's AI agent joins the conversation automatically, no setup required. Record, transcribe, and analyze in real-time, all from your browser.

Chrome extension: Automatic meeting recording

Install the Swiftask extension from the Chrome Web Store and meetings start recording instantly. When you open Google Meet or Teams, a popup appears offering to record. Select your AI agent, approve its entry to the meeting, and watch it transcribe in real-time, no manual clicking, no switching between apps.

• One-click setup – Install from Chrome Web Store, then meetings auto-detect

• Smart agent selection – Choose which AI agent handles transcription for each meeting

• Bot joins automatically – Agent enters as a participant (with your approval) and captures everything

• Real-time transcription – View the transcript as the meeting happens in your web interface

• Works with – Google Meet, Teams, and other major platforms

How to test it:

1. Install the Swiftask extension from the Chrome Web Store

2. Open Google Meet or Teams in your browser

3. The popup appears automatically, select your preferred AI agent

4. Approve the agent's entry to the meeting

5. Watch the real-time transcription appear in your Swiftask web interface

6. Access summaries, action items, and decisions just like desktop recordings

Web interface: Monitor meetings in real-time

Can't use the extension? Copy your meeting link and paste it into Swiftask's web interface. The AI agent joins the same way, and you get the same real-time transcription and analysis, all without leaving your browser.

• Paste meeting link – No installation needed; just copy and paste the meeting URL

• Real-time viewing – Watch transcription appear as the meeting progresses

• Full analysis – Extract action items, summaries, and decisions instantly

• Available on all plans – Pro, Team Starter, Team Growth, and Custom

How to test it:

1. Copy your meeting link (Google Meet, Teams, etc.)

2. Go to Meetings → Record Meeting in Swiftask web

3. Paste the meeting link in the interface

4. Select your AI agent and approve its entry

5. See real-time transcription in the web interface

6. Use all standard meeting analysis features

See how it works in this video:

Key details

• Desktop app still powers automatic detection – The Swiftask desktop app continues to auto-detect meetings on your computer and record without any action

• Agent validation required – Meeting participants see the bot request entry; you approve it to start transcription

• Same analysis everywhere – Whether recorded via extension, web link, or desktop app, all meetings get the same AI-powered summaries, action items, and decision tracking

• Available on all plans – No plan restrictions; everyone gets browser-based meeting recording

Ready to try it? Install the Swiftask extension from the Chrome Web Store, open your next meeting, and let the AI handle the note-taking.

Turn your Python functions into reliable agent skills. Define once, execute infinitely.

Until now, you had two choices: either let your agents generate Python code on-the-fly (flexible, but unpredictable), or manually write every integration. Neither was ideal.

You had solid Python code; tested code, code that works, but your agents couldn't use it directly.

We've created Python Code Execution: a way to turn your Python functions into reusable agent skills.

Here's how it works:

1. You write a Python function (a real one, with your business logic)

2. You deploy it as an agent skill in 2 minutes

3. Your agents execute it whenever they need it, always the same way

No dynamic generation. No hallucinations. No surprises. Just your code.

What You Can Do Now

• Currency conversion: Your agents convert amounts with live exchange rates, without asking you

• Data validation: Normalize emails, phone numbers, addresses before they touch your database

• Complex business logic: Calculate discounts, taxes, fees, exactly as your rules require

• API integration: Wrap your API calls with authentication and error handling, then let agents use them

• Data transformation: Clean, reformat, enrich data before sending it elsewhere

• Structured reports: Aggregate and format data for presentation or automatic export

Technical Details

Your code runs in an isolated, secure environment:

• Timeout: 30 seconds per execution (configurable by plan)

• Memory: 512 MB per execution

• Libraries: You have access to requests, pandas, numpy, json, datetime, and more

• Errors: Smart handling, errors are captured and returned usefully

• Traceability: Every execution is logged for audit and debugging

Get Started in 5 Minutes

1. Open an agent (existing or new)

2. Go to Skills → Add a skill

3. Select "Python Code Execution"

4. Paste your Python function (e.g., a currency converter)

5. Test with real data

6. You're done.

Concrete Example:

def convert_currency(amount: float, from_currency: str, to_currency: str) -> str: import requests url = f"https://api.frankfurter.app/latest?amount={amount}&from={from_currency}&to={to_currency}" response = requests.get(url, timeout=10) data = response.json() converted = data["rates"][to_currency] return f"{amount} {from_currency} = {converted} {to_currency}" 

Save, test, it's live.

Python Code Execution vs. Python Code Interpreter : When to Use Which?

We also offer Python Code Interpreter (dynamic code generation). Here's the difference:

Use Code Execution if:

• You have code that works and needs to stay stable

• You need total control over the logic

• It's for production (payments, sensitive data, etc.)

Use Code Interpreter if:

• You're exploring, experimenting

• Every request is different

• You don't have predefined code

Availability

• Pro plan and higher (Team Starter, Team Growth, Custom)

• Not available on the free plan

• Where to find it: Agents → [Your agent] → Edit → Skills → Add a skill → "Python Code Execution"

What This Changes for You

Before: You had Python code + your agents = two separate worlds.

Now: Your Python code = an agent skill = a complete, end-to-end workflow.

No more need to:

• Rewrite your business logic

• Ask your agents to generate risky code

• Maintain two versions of the same logic

Just: Define once, execute infinitely.

"Ready to turn your Python code into an agent skill? Check out the documentation."

→ Discover at docs.swiftask.ai

We've launched the Datagouv Model Context Protocol (MCP) server, connecting Swiftask agents directly to France's largest open data platform. Your agents can now query public datasets on business information, legal documents, real estate, employment, and more, without switching tools or writing custom connectors.

Datagouv MCP server integration

French enterprises now have direct access to France's most comprehensive open data repository through a dedicated MCP server. This means your agents can instantly retrieve and analyze public datasets across multiple domains: from company registry data (INSEE/Sirene) to real estate valuations (DVF), employment statistics (DARES), and public tender information (BOAMP).

The integration uses the Model Context Protocol (MCP) standard, Swiftask's new connector architecture that eliminates the need for manual API configuration. Simply enable the Datagouv skill on your agent and start querying French public data in natural language.

Supported data sources:

• Company information: INSEE Sirene API (business registry, SIREN/SIRET lookups)

• Legal & compliance: Légifrance API (French legislation, regulations)

• Public tenders: BOAMP API (French government procurement notices)

• Real estate: DVF (real estate transaction data)

• Employment: DARES open data (labor market statistics)

• Health & social: Ameli/HAS data (healthcare information)

• Macro-economic: INSEE/Eurostat APIs (national statistics)

• EU regulations: EUR-Lex (European legal documents)

How to use it:

1. Open your agent configuration → Skills section

2. Search for and add the Datagouv MCP server skill

3. Configure access mode (read-only recommended for public data)

4. Start asking your agent questions like: "What companies are registered in Lyon?" or "Show me recent public tenders for construction services"

How to test it:

1. Create or open an existing agent

2. Add the Datagouv MCP server skill from the Skills library

3. In Chat, ask your agent to retrieve data: "Find all businesses registered to [company name]" or "What are the latest public procurement notices?"

4. Verify responses include accurate data sourced directly from Datagouv

5. Test filtering and sorting capabilities with natural language queries

Availability: All plans (Pro, Team Starter, Team Growth, Tailored)

What this enables:

• Compliance teams can instantly verify company information and regulatory requirements

• Business analysts can access economic data and labor market trends without manual lookups

• Government relations teams can monitor public tenders in real time

• Real estate professionals can query transaction data and valuations instantly

• Researchers and consultants can combine Datagouv data with your own knowledge base for richer analysis