Changelog

Questions about knowledge tables? We've got you covered.

What are knowledge tables?

Knowledge tables let you organize your business data (leads, products, FAQs, customer info, test cases) in structured, queryable tables. Your AI agents can instantly access this data to provide accurate, context-driven responses—no manual lookups needed.

Think of them as: Organized spreadsheets your agents can understand and use automatically.

How do they work?

1. Create a table with custom columns (text, numbers, dates, booleans, and more)

2. Add your data

3. Connect the table to your agents

4. Your agents query the data in real time to answer questions or complete tasks

Available on all plans

Knowledge tables work on Pro, Team Starter, Team Growth, and Custom plans—no extra cost, no restrictions.

Learn more

For step-by-step guides, use cases, and technical details, visit our documentation:

🔗 Knowledge tables — Complete guide

Have questions? Our support team is here to help.

We're thrilled to introduce Knowledge tables, a new way to structure and organize your business data so your AI agents can access it instantly. No more scattered spreadsheets, keep your leads, products, FAQs, and customer information in one place, ready for your agents to use.

Knowledge tables

Create simple, organized tables for any type of business data, then connect them to your agents. Whether you're managing leads, product catalogs, or FAQ databases, knowledge tables let you structure information in a way that makes sense for your team and your AI.

• Create tables with custom columns and rows, no coding required

• Organize leads, products, FAQs, or any structured data you need

• Agents can query and access your tables to provide accurate, data-driven responses

• Available immediately in the Knowledge section: Knowledge → Knowledge tables

• Search tables by name using the built-in search bar

• No technical limits, create as many tables as you need

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

How to test it: Navigate to Knowledge in your sidebar, click the Knowledge tables tab, then click "+ New table" to create your first table. Add columns, enter your data, and immediately make it available to your agents.

What this means for you

For team leads: Stop managing data across multiple tools. Keep your leads, products, and FAQs in Swiftask where your agents can access them in real time.

For agents: Your AI now has direct access to structured, up-to-date business data, enabling more accurate and contextual responses.

For everyone: Knowledge tables are available to all users on all plans, no restrictions, no extra cost, no setup complexity.

Getting started

1. Go to Knowledge → Knowledge tables

2. Click + New table

3. Name your table and add columns (text, numbers, dates, booleans, and more)

4. Add your data

5. Your agents can now query the table and use the data in their responses

Ready to organize your data? Start creating your first table today!

We have integrated five new LLM models into Swiftask to strengthen your AI agnosticism and offer you even more options tailored to your needs. Discover xAI Grok 4.3, Kimi K2.6, Qwen3.6 Plus, Qwen3.6 Flash, and Qwen3.6 Max Preview, all available immediately, at no extra cost.

xAI Grok 4.3

Grok 4.3 brings deep understanding and fast reasoning. Designed for analytical tasks and workflows requiring nuanced reflection, this model excels in analyzing complex data and solving multi-step problems.

• Ideal for: Strategic analysis, logical reasoning, processing complex documents

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

• Access: Chat → Model Selector | Agents → Configuration → Model Selection

Kimi K2.6

Kimi K2.6 offers low latency and fine-tuned contextual understanding. Perfect for real-time interactions and tasks requiring quick responses without compromising on quality.

• Ideal for: Interactive chat, real-time transcription, quick answers

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

• Access: Chat → Model Selector | Agents → Configuration → Model Selection

Qwen3.6 Plus

Qwen3.6 Plus combines performance and versatility. This balanced model is suitable for a wide range of tasks, from writing to analysis and generating structured content.

• Ideal for: General writing, summaries, content analysis, artifact generation

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

• Access: Chat → Model Selector | Agents → Configuration → Model Selection

Qwen3.6 Flash

Qwen3.6 Flash is optimized for speed. It is ideal for high-volume tasks and workflows where responsiveness takes precedence over analytical depth.

• Ideal for: Repetitive tasks, support chat, bulk processing, quick interactions

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

• Access: Chat → Model Selector | Agents → Configuration → Model Selection

Qwen3.6 Max Preview

Qwen3.6 Max Preview is the most powerful version of the Qwen3.6 series. Designed for demanding tasks, it offers advanced reasoning and expanded multimodal capabilities.

• Ideal for: Complex reasoning, in-depth analysis, critical tasks, strategic problem-solving

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

• Access: Chat → Model Selector | Agents → Configuration → Model Selection

How to access these new models

In Chat:

1. Open Chat

2. Click the Model Selector at the top of the input field

3. Browse the list and select one of the new models

4. Start conversing

For your Agents:

1. Go to Agents

2. Select your agent and click Edit

3. Open Configuration → Model Selection

4. Choose one of the new models

5. Save

For Default Settings:

1. Go to Workspace administration → Settings → Advanced

2. Select your default model for Chat, Meeting AI, Transcription AI, or Artifact AI

3. Click Update

How to test each model

xAI Grok 4.3:

• Ask a strategic question: "Analyze this business problem and propose three solutions with their pros and cons"

• Test chain-of-thought reasoning on complex logical tasks.

Kimi K2.6:

• Measure response speed with a simple question.

• Example: "Summarize this article in 3 key points [paste content]"

• Ideal for testing in fast chat mode.

Qwen3.6 Plus:

• Try a writing task: "Draft a professional email for [context]"

• Test analysis: "Extract the key points from this document"

Qwen3.6 Flash:

• Submit multiple questions in a row to measure responsiveness.

• Test on repetitive or classification tasks.

Qwen3.6 Max Preview:

• Upload a large document and request a detailed analysis.

• Try a multi-step reasoning task.

• Example: "Analyze this report and propose a data-driven strategy"

Improvements

Real-time AI model recommendations now include all five new models. Simply describe your task in the chat, and Swiftask will automatically suggest the best-fitted model, no manual selection required.

All these models are included in your current plan. No upgrade necessary.

Open Chat now and test these new models to discover the one that will transform your workflow. 🚀

Important Notes

• ✅ All models are available on all plans (Pro, Team Starter, Team Growth, Custom)

• ✅ No extra cost, included in your current subscription

• ✅ Switch between models in 1 click according to your needs

• ✅ Compatible with your existing agents, Knowledge Base, and skills without modification

Ready to explore this new generation of models? Open Chat now and test each one. 🎯

You can now create and manage AI agents directly from your application using our REST API. Instead of configuring agents through the interface, send a single API request with your agent's name, description, system prompt, and greeting messages. Your agent is created, configured, and ready to use instantly.

This is perfect for platforms integrating Swiftask, SaaS products offering white-label agents, or teams managing large agent fleets across different projects.

Create Agent API

The new Create Agent endpoint lets you automate agent deployment. Define all required properties (name, description in English and French, system prompt, greeting messages), send a POST request to https://api.swiftask.fr/admin/agent/create, and receive your agent's ID, slug, and status in response.

• Endpoint: POST https://api.swiftask.fr/admin/agent/create

• Required fields: 6 mandatory properties (name, description, descriptionFR, systemPrompt, greetingMessage, greetingMessageFR)

• Field constraints: Agent name up to 100 characters; description up to 500 characters

• Response: Agent ID, slug, status, and creation timestamps

• SDK compatible: Works with OpenAI SDK (Python, JavaScript, TypeScript, Node.js)

• Streaming: Full support for real-time responses via Server-Sent Events (SSE)

• Available on: All paid plans (Pro, Team Starter, Team Growth, Tailored) with no usage limits

• Find it in: Agent Settings → API tab

How to test it:

1. Go to Account Settings → API → Create new key. Copy it immediately (displayed only once).

2. Open your agent → Agent Settings (gear icon) → API tab → Copy your agent slug.

3. Install the OpenAI SDK: pip install openai (Python) or npm install openai (JavaScript).

4. Use this Python example to test:

   from openai import OpenAI client = OpenAI(api_key="YOUR_API_KEY", base_url="https://api.swiftask.fr/v1") response = client.chat.completions.create( model="your-agent-slug", messages=[{"role": "user", "content": "Test message"}] ) print(response.choices[0].message.content) 

5. Verify the agent responds using its configured knowledge base and skills.

Use cases:

• Platform integration: Embed Swiftask agent creation into your SaaS product

• White-label agents: Let your customers deploy branded AI agents without accessing Swiftask directly

• Fleet management: Automate agent creation and configuration across your organization

• Custom workflows: Build agents programmatically as part of your business logic

Full compatibility, zero friction

Swiftask agents are now fully compatible with the OpenAI SDK. You can use any OpenAI client library (Python, JavaScript, Node.js, or others) to send messages to your agents and receive responses, just as you would with OpenAI's own models. No special integration code needed, just point the SDK to Swiftask's API endpoint, authenticate, and go.

OpenAI SDK Compatibility for Agents

Your Swiftask agents now work seamlessly with the OpenAI SDK. This means you can leverage the familiar OpenAI interface, the one your team already knows, while powering it with your custom agents, knowledge bases, and skills.

Whether you're integrating agents into existing applications, automating document processing, or building real-time chat experiences, the OpenAI SDK compatibility makes it trivial to do.

• Base URL: https://api.swiftask.fr/v1

• Authentication: Use your existing Swiftask API key (generate one in Account Settings → API)

• Model parameter: Specify your agent's slug (found in Agent Settings → API tab)

• Supported SDKs: Python, JavaScript, TypeScript, Node.js, and any other OpenAI-compatible client

• Streaming: Full support for real-time responses via Server-Sent Events (SSE)

How to test it:

1. Create an API key : Go to Account Settings → API → Create new key. Copy it immediately (shown only once). Optionally set an expiration date for security.

2. Get your agent's slug : Open your agent → Settings (gear icon) → API tab → Copy the Agent slug.

3. Install the OpenAI SDK : Run pip install openai (Python) or npm install openai (JavaScript).

4. Configure and call your agent : Here's a quick Python example:

from openai import OpenAI client = OpenAI( api_key="YOUR_API_KEY", base_url="https://api.swiftask.fr/v1" ) response = client.chat.completions.create( model="AGENT_SLUG", messages=[{"role": "user", "content": "Hello, how can you help me?"}] ) print(response.choices[0].message.content) 

And the same in JavaScript:

import OpenAI from 'openai'; const client = new OpenAI({ apiKey: 'YOUR_API_KEY', baseURL: 'https://api.swiftask.fr/v1', }); const response = await client.chat.completions.create({ model: 'AGENT_SLUG', messages: [{ role: 'user', content: 'Hello, how can you help me?' }] }); console.log(response.choices[0].message.content); 

5. Enable streaming (optional) : For real-time responses, set stream=True (Python) or stream: true (JavaScript). The response will stream back character by character, giving users immediate feedback.

Practical Use Cases

Integrate agents into support chatbots : You have a customer support agent in Swiftask and want to use it in your existing Node.js application. Now you can use the OpenAI SDK directly, no custom code needed.

Batch process documents : Process hundreds of documents with your document analysis agent using a simple Python loop. The agent processes each one and returns structured results.

Stream responses in web apps : Display agent responses in real-time on your web page. Streaming makes the experience feel faster and more interactive.

Best Practices

• Secure your API key : Treat it like a password. Store it in environment variables or secret management tools, never hardcode it.

• Set expiration dates : When creating an API key, set an expiration date for security. Rotate keys periodically.

• Use descriptive key names : Name your keys by use case (e.g., "Production Bot", "Testing Bot") to track usage and manage access.

• Handle errors gracefully : Wrap API calls in try-catch blocks and provide meaningful error messages to users.

• Use streaming for better UX : Streaming responses feel faster to end users because they see content appear in real-time instead of waiting for the full response.

Get Started Now

Ready to use your agents with the OpenAI SDK?

Head to Account Settings → API, create your key, grab your agent's slug, and start building.

Your agents are now as easy to integrate as OpenAI's GPT models. 🎯

More informations : https://docs.swiftask.ai/

We are thrilled to announce the addition of two powerful models to the Swiftask platform. OpenAI GPT-5.5 brings advanced reasoning capabilities, while GPT Image 2 revolutionizes image generation with near-perfect text rendering and photorealistic quality. Both models are now available on all plans.

OpenAI GPT-5.5 — Your Most Advanced Reasoning Engine

GPT-5.5 is OpenAI's latest model, designed for tasks where precision and depth are paramount. It replaces GPT-5.4 and offers superior performance on complex problems and nuanced instructions.

• 1.1M Token Context Window: Process large documents, complex code, and long conversations in a single request.

• Enhanced Reasoning: Greater accuracy for complex logic, analysis, and research.

• Better Instruction Following: Fewer surprises in extended interactions.

• Multimodal Features: Integration of vision, web search, and function calling.

• Available on All Plans: Pro, Team Starter, Team Growth, and Custom at no extra cost.

OpenAI GPT Image 2 — Image Generation Reimagined

GPT Image 2 is a complete overhaul, delivering sharper images and exceptional text rendering. Launched in April 2026, it sets new standards for AI-generated imagery.

• Near-Perfect Text Rendering: Over 95% accuracy in multiple languages, including English, French, Japanese, and Chinese.

• Accurate Colors: Elimination of yellow tints, with neutral and realistic colors.

• Flexible Resolutions: Image generation from 1024×1024 up to 4K (3840×2160).

• Three Quality Levels: Low for quick drafts, Medium for everyday use, and High for final assets.

• Contextual Understanding: Image generation that accounts for meaning and spatial relationships.

How to Access These New Models:

• In Chat: Click the AI model selector → Select "OpenAI GPT-5.5" or "OpenAI GPT Image 2".

• For Agents: Agent configuration → Model selection → Choose the desired model.

• For Sub-Agents: Sub-agents configuration → Model selection → Choose the desired model.

With these additions, you no longer need to juggle multiple tools or subscriptions. Everything you need is now available on a single platform, accessible to your entire team. Try these new features now!

Managing multiple agent conversations just got easier. We've added smart filtering to your chat history, and conversations now stay organized the way you work.

Filter conversations by agent

You can now quickly find conversations by agent directly in your chat history. Use the "Filter by agent" button at the top right of your search field to filter by agent, or search by agent name to instantly find the conversations you need.

• Find it in: Chat Hub (left sidebar) → Search field → "Filter by agent" button

• Search by: Agent name or use the dropdown filter

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

How to test it:

1. Open the Search field in the Chat Hub from your left sidebar

2. Look for the "Filter by agent" button at the top-right of your search field

3. Click the dropdown or use the search field to type an agent name

4. Your conversation list instantly filters to show only that agent's sessions

Active conversations stay at the top

Chat sessions now automatically move to the top of your list whenever a new message is sent. Your most active conversations are always within reach, keeping your focus on what matters.

• Enabled by default for all users

• No configuration needed, it just works

How to test it:

1. Open a chat conversation from your history

2. Send a message to the agent

3. Return to your chat history, the session should now appear at the top of the list

4. Send messages to different agents and watch the list reorder automatically

We've shipped five powerful improvements that give your agents more intelligence, more control, and better efficiency. From thinking deeper to routing smarter, here's what's new.

Your agents can now think deeper

Reasoning effort configuration lets you control how much analysis your agent performs before responding. Set the thinking level to Low, Medium, or High to balance speed, quality, and credit usage.

• Low: Fast responses, minimal analysis, perfect for simple questions and quick replies

• Medium: Balanced analysis, ideal for most tasks

• High: Deep reasoning, best for complex problems, detailed analysis, and nuanced decisions

Higher thinking levels increase token consumption per message, so your credits are used faster. This is particularly useful for advanced reasoning tasks where depth matters more than speed.

Find it at: Agent settings → Agent Instruction → LLM → Reasoning effort

Compatible with: Anthropic Claude models (4.5, 4.7 series) and equivalent advanced reasoning models

Available on: All plans

How to test it: Open an agent, set reasoning effort to High, ask a complex analytical question, and observe the deeper reasoning in the response. Then compare with Low setting to see the difference.

Chat just got cleaner and more focused

Improved chat welcome interface is now centered and more immersive, making it easier to navigate between agents and access pinned agents at a glance. Perfect for admins who want to pin their most-used agents while changing the default agent to something other than Swiftask.

The centered layout eliminates sidebar clutter and puts your agents front and center, so switching between specialized agents (content writer, data analyst, customer support) is faster and more intuitive.

Available on: All plans

How to test it: Go to Chat, pin 2–3 of your favorite agents, then change the default agent. Notice how the pinned agents are now more visible and easier to access.

Let your agents pick the right LLM for the job

Auto mode (beta) is intelligent LLM routing that lets you define multiple models and automatically routes each user question to the best model for the task. Save credits by using economical models for simple questions and advanced models only when needed.

How it works:

• Define multiple LLM options in your agent configuration

• Set routing criteria based on question complexity and type (e.g., "Use Gemini Flash for simple questions, Claude for analysis")

• The system analyzes each user query and automatically selects the most cost-effective model

• Users can pin a specific model for the session if they prefer consistency

The routing is transparent and automatic, no special prompts needed. But you can influence behavior by specifying task priorities in your agent instructions.

Find it at: Agent settings → Agent Instruction → LLM auto mode settings

Example: Route simple FAQ questions to a fast, economical model (saving credits), while routing complex analysis to a powerful model only when needed.

Available on: All plans

How to test it: Enable Auto mode, define 2–3 models with different rules, then ask your agent a simple question and a complex one. Watch as it routes each to the appropriate model.

Your agents can now create interactive diagrams

Diagram creation in Artifacts lets your agent generate flowcharts, sequence diagrams, mind maps, and other visual structures directly in your artifacts. All diagrams are fully editable : modify colors, text, layout, and connections right in the artifact after generation.

Perfect for:

• Visualizing processes and workflows

• Creating org charts and hierarchies

• Mapping out project timelines and dependencies

• Documenting system architecture

Your agent detects visualization requests in your prompts and automatically generates structured diagrams. The diagrams render in an interactive, editable format (Mermaid-style), so you can tweak them without regenerating.

Find it at: Artifacts → Create new diagram → Choose an AI agent

Available on: All plans

How to test it: Ask an agent "Create a flowchart of our customer onboarding process" and watch it generate an editable diagram. Then click on the diagram to modify colors, add steps, or reorganize the flow.

Email management just got smarter

Enhanced IMAP/SMTP email handling improves how your agents manage email workflows. Sent folders are now synchronized bidirectionally, and threading keeps conversation context intact throughout long email exchanges.

What's improved:

• Sent folder sync: Your sent emails are automatically synchronized between Swiftask and your email client

• Threading: Long email conversations stay grouped together, preserving context and preventing lost information

• Seamless integration: Works with Gmail, Outlook, and other IMAP/SMTP providers

• No impact on triggers: Agent email triggers (Trigger from Outlook, Trigger from IMAP) continue to work exactly as before

This is especially useful for agents handling customer support, sales follow-ups, or any workflow where email context matters.

Available on: All plans

How to test it: Set up an agent with email skills, send a series of emails back and forth, and verify that the conversation thread stays intact and sent emails appear in both Swiftask and your email client.

Tips & best practices

For reasoning effort: Start with Medium for most tasks. Use High only for complex analytical work where depth matters more than speed, since it consumes more credits.

For Auto mode: Define clear, simple rules (e.g., "Use Gemini Flash for questions under 50 words, Claude for longer analysis"). Test with real questions to ensure routing matches your expectations.

For chat navigation: Pin your 3–5 most-used agents. This keeps your workspace focused and makes switching between specialized agents instant.

For diagram creation: Use specific language in your prompts ("Create a flowchart of...", "Show me the sequence diagram for..."). The more structured your request, the better the diagram.

For email workflows: Always enable threading in your email skills to preserve conversation context. This prevents miscommunication and keeps your audit trail clean.

Ready to upgrade your agents? Start with Auto mode to optimize credit usage, then use reasoning effort for your most complex tasks. Your agents are now smarter and more efficient.