The latest updates to the Younet AI platform.

Version – v. 9.6                                       

  (May 6, 2026)

Features & updates:

• Custom MCP Servers (Beta) – Users can now bring their own MCP (Model Context Protocol) servers into Younet, dramatically expanding what agents can do beyond our built-in toolset. A curated template gallery makes setup instant for popular providers like Zapier, Atlassian, Notion, HubSpot, Asana, Canva, monday.com, Intercom, Linear, Pipedream, Square, PayPal, Stripe, GitHub, and Cloudflare — or users can plug in any custom remote MCP server of their own. A new live capabilities preview lets users see exactly what tools and resources a server exposes before activating it. Backend MCP support has also been refreshed to match the latest spec for full compatibility.
• Modern OAuth 2.1 for MCP – Connecting MCP servers now uses the latest OAuth 2.1 standard with automatic client registration, so most integrations can be connected with a single sign-in popup — no developer setup required. For providers that need it, a manual fallback lets users supply their own credentials.
• 3,000+ Connectors – A massive expansion of the integrations marketplace. Agents can now connect to over 3,000 third-party apps through our Pipedream partnership, covering Sales, Marketing, Project Management, Finance, Design, HR, Support, Analytics, and beyond. The new Integrations Marketplace page offers a searchable catalog with category filtering and a unified grid view. Users connect an account once and any of their agents can use the related actions during a conversation.
• Unified Connector Experience – In-house tools and Pipedream apps are now organized side-by-side under a single, consistent category structure. Categories that previously appeared twice (e.g., "Analytics" vs. "Data Analytics") are intelligently merged, duplicates between in-house and Pipedream versions of the same service are hidden, and active integrations automatically rise to the top of each category for quicker access.
• Smarter Bulk Enable for Categories – Toggling a whole category of connectors now opens a clear confirmation dialog showing exactly how many tools will be activated and how many still need to be connected first — no more accidental mass-toggles.
• Stop Button for In-Flight Responses – Users can now stop an AI response mid-stream with a single click. The visible portion of the response is preserved with a "(stopped by user)" marker, the conversation continues cleanly, and billing reflects only what was actually generated. Works across streaming responses, tool execution, and stop+resend cycles, and propagates instantly across all supported providers (OpenAI, Claude, Gemini).                                 
• Faster Streaming Responses – Removed a 6–10 second lag that some users saw at the end of long replies, and reduced backend overhead on every streamed chunk. Resumed streams (e.g. after a brief disconnect) no longer hang before continuing.   
• Improved Stability Under Heavy Load – Reworked how the platform tracks in-flight chat requests under concurrent load, eliminating a backend bottleneck that could cause app-wide slowdowns when many users were chatting at once.
• Smarter Video Generation Credits – Replaced the upfront credit reservation for video generation with a just-in-time check. Users with video tools available no longer get pre-rejected when they have no intent to use them, and credits are only held when generation actually starts.
• Public Agent Discovery – Guest visitors can now browse public agent metadata without authentication, making shared and public agents easier to preview before signing up.
• Improved Tool Calling Across Providers – Tools and web search now work consistently across OpenAI's Responses API, Claude, and Gemini reasoning models, so agents behave the same regardless of the underlying model.
• Repositioned Agent Settings & Renamed Deploy Button Agent Settings have been moved from the "Brain" section directly to the main menu for faster access during chat sessions, and the primary action button has been renamed from "Deploy Agent" to a more concise "Deploy."
• Vendor-Neutral Web Search — Removed the OpenAI badge from web search results for a cleaner, model-agnostic experience search results now look the same regardless of which underlying model is in use.
• Refreshed Source & Connector Icons — Updated iconography across the connector picker and source attribution for better visual clarity and consistency with the rest of the platform.
• Polished "Thinking" Indicator — Refined the loading animation shown while an agent is thinking for a more subtle, less distracting feel.

New AI Models Added:;

• Claude Opus 4.7 (Anthropic, Apr 16) — Anthropic's most capable model to date, with 1M token context, excelling at agentic coding, long-horizon tasks, and high-resolution vision.
• Claude Sonnet 4.6 (Anthropic, Feb 17) — Near-Opus performance at Sonnet pricing.
• Claude Opus 4.6 (Anthropic, Feb 5) — Previous-gen flagship with adaptive thinking and 1M context.
• GPT-5.4 & GPT-5.4 Pro (OpenAI, Mar 5) — OpenAI's new flagship tier with state-of-the-art coding and a 1M token context window.             
• GPT-5.4 Mini & Nano (OpenAI, Mar 17) — Faster, cost-efficient variants for everyday and high-volume tasks.                         
• Gemini 3.1 Pro (Google, Feb 19) — Google's most advanced reasoning model, replacing the retired Gemini 3 Pro.               
• New OpenAI Image Generation Model — Added with extended generation parameters for higher-quality image output.             

Bug fixes:

• Playground Typing Lag – Resolved a significant typing slowdown that occurred in long chats with large tool-result histories. Typing in the playground input is now smooth even after dozens of agent tool calls.     
• PDF Knowledge Base Stability – Resolved 500 errors users encountered when ingesting certain PDF files into an agent's brain.
• Silent Drops of Large DOCX Files on Claude Models – Fixed an issue where large Word documents (22k+ tokens) sent to Claude 4.5 / 4.6 / 4.7 were silently dropped, causing the agent to respond as if no file had been attached. File-size limits have been recalibrated to each model's actual capacity.                       
• Gemini Tool & MCP Compatibility – Fixed schema-resolution issues that caused "Something went wrong" errors on Gemini 3 Pro when using certain MCP tools. A single malformed tool definition no longer crashes the entire request. 
• Reasoning Output for Latest Models – Restored visible reasoning steps in chats using Claude Opus 4.7 and Gemini 3.1.                       
• Pre-Made Agent Lock Carryover – Fixed a bug where selecting a pre-made agent (e.g., Social Media Manager) caused its read-only configuration lock to persist when the user switched back to their own personalized agents.   
• LLM Looping on Low Balance – Fixed an issue where users near their credit limit would see the model repeatedly retry video generation after each rejection. The agent now responds with a single clear message and stops retrying.             
• Connector Lock Indicator – Fixed a UI issue where some connectors incorrectly displayed a "locked" state.                       
• Public/Shared Agent Connectors – Visitors using a shared or public agent now correctly see the agent owner's connected tools, fixing a regression where guest sessions silently lost connector access.     
• Batch Status Toggle – Fixed an issue where toggling status on multiple agents at once would silently drop some of the changes.                     
• Video Generation Billing – Removed an upfront credit reservation that could leave credits stuck if a video request was abandoned. Charges now happen only when generation actually starts.           
• Paywall for New Users – Corrected a paywall flow that was incorrectly shown to certain new accounts.                                                   
• Removed Misleading Error Popup – Removed an error popup that surfaced too aggressively on transient network conditions.             
• OpenAI Search Bug – Fixed an issue affecting OpenAI model search results.                                                                 
• MCP Dark Mode – JSON capability panels now respect the active theme and render correctly in dark mode.   
• MCP Remote-Only Clarification – Updated UI copy to clearly indicate that only remote (HTTP/SSE) MCP servers are supported — local stdio servers are not. 

Version – v. 9.5

(April 17, 2026)

Features & updates:

• Connector Icons on Agent Cards – Agent cards in the marketplace now display connector icons, making it easy to see which tools and integrations are attached to each agent at a glance. Hover over the icons to view connector names.

• Video Generation Tool – Video generation is now available as a first-class tool in the marketplace, allowing agents to create videos directly within conversations.

• Guest Registration Prompt – Guest users will now see a one-time registration prompt after sending their first message to an agent, making it easier to create an account and continue the conversation.

• Improved Billing Accuracy – Token usage and video generation costs are now tracked more precisely, ensuring billing reflects actual usage across all supported models.

Bug Fixes

• Duplicate Image Attachments – Fixed an issue where attaching two images in a chat would incorrectly show three attachments, and images would re-attach on subsequent messages.

• PDF Processing – Improved reliability of PDF file processing for knowledge base ingestion.

Version – v. 9.4

(March 19, 2026)

Features & updates:

• Integrated Google Account Connectivity – Users can now connect Google accounts directly from the tools panel for Gmail and Calendar. Unconnected tools now display a "Sign in with Google" button rather than disabled toggles, unlocking automatically upon successful connection.
• Streamlined OAuth Popup Flow – Implemented a popup-based authentication window that allows users to sign in without losing their current page context. The panel updates instantly and the window closes automatically after completion.
• Simplified Third-Party Integrations – Users can now authenticate at the provider level (e.g., Google, Slack) once, granting access to all related functions within their model rather than requiring separate authentication for each individual function.
• Enhanced Request Structure – Updated the system to support provider-based authentication by accepting provider IDs directly during OAuth callbacks, enabling a more seamless connection experience.
• Linked Agent Privacy Dialogue – Added an informational popup that appears when a user attempts to select files from a linked agent's brain. The dialog explains privacy restrictions and suggests alternatives, such as uploading temporary files or creating a personal agent.

Bug fixes:

•  Backend Security & Stability Upgrade – Replaced the legacy session-based authentication flow with a reliable stateless approach. This resolved the persistent 500 errors users encountered when connecting to Google services.

Version – v. 9.3

(March 16, 2026)

Features & updates:

• Redesigned the connectors popup by introducing groups and subgroups for better organization, including categorizing existing default connectors.
• Enhanced the connectors popup search functionality to include group names, and added counters to display the number of selected connectors at both the group and subgroup levels.
• Overhauled the prompt examples/suggestions block to anchor it directly to the chat input area so it remains visually attached regardless of the scroll state.
• Updated the UI to hide the "Show" button for prompt examples when no active agent context is selected.
• Collected the task list from the team meeting and created the corresponding tasks in the CRM.

Bug fixes:

• Fixed a visual blur issue on the "for fee" preview page.

Version – v. 9.2

(March 12, 2026)

Features & updates:

• Rewrote the connectors logic to align with standard button behaviour, updating all related components across the playground, shared models, and landing page.

• Renamed the "Share" button to "Deploy Agent" and updated its icon to a rocket.

• Updated the styling for the footer and personalized block to meet the latest design requirements.

• Added a new toggle feature to show or hide prompt examples directly within the playground chat screen.

Bug fixes:

• Fixed a UI bug that caused the trial modal to appear in the incorrect position.
• Resolved a z-index layout issue with the scroll-down button.

• Fixed an issue that prevented CSV files from being uploaded via the brain select dropdown in the playground chatbox.

• Corrected styling issues with the dark/black theme on the "upload file to brain" pop-up.

• Updated frontend and backend domain validation rules to resolve an issue that blocked certain valid domains from being added for parsing.

• Fixed a UI bug where users with paid accounts were incorrectly shown a premium upgrade indicator (purple star) when selecting LLMs after clicking the "Regenerate" button.

Version – v. 9.1

(February 26, 2026)

What’s New / Improvements

1. UI & Home Page Redesign (AI Agents-first)

• Updated overall interface and home page layout to make AI agent discovery and selection faster and more intuitive.

2. Connectors & AI Agent Selection UX Refresh

• Redesigned Connectors flow and AI agent selection experience to reduce friction and improve clarity across the setup process.

Fixes

3. Multiple UI Bug Fixes

• Resolved a set of UI issues to improve stability, consistency, and overall user experience.

Version – v. 9.0

(January 12, 2026)

We're thrilled to announce a major enhancement to the Younet AI platform — a brand-new LLM Selector Menu now available directly within the input field of all AI Agents and AI Models.

This update empowers you with the flexibility to choose your preferred large language model on the fly, giving you full control over the intelligence powering your workflows.

Available Models

Key Benefits

• Flexibility – Switch between models instantly based on task complexity, speed, or cost.
• Performance Optimization – Use lightweight models for quick tasks and advanced models for deep reasoning.
• Seamless Experience – Integrated directly into the input interface — no additional setup required.

Version – v. 8.8

(October 9, 2025)

Introducing the Phone Agent (BETA)

Turn your agent into a 24/7 phone assistant with an all-new, tabbed management experience.

Transcripts Tab

• Play & follow along: Listen to call audio while the transcript auto-scrolls in sync with timestamps.
• Full call history: Smooth browsing with pagination to review past conversations.
• Download recordings: Save audio files for hand-offs or audits.

Analytics Tab

• At-a-glance insights: See call volumes, durations, outcomes, and trends in clean visualizations.
• Spot patterns: Identify peak times and performance shifts to tune your agent.

Voice & Number Management

• Pick a voice: Choose from a gallery and preview instantly.
• Get a number fast: Purchase or rotate your Phone Agent’s number without leaving the page.

Version – v. 8.7.1

(August 28, 2025)

Bug Fixes & Technical Improvements

• Long responses interruptions: The long responses from AI agents' interruptions have been fixed.

Version – v. 8.7

(July 31, 2025)

Introducing the All-New Social Media Manager Agent

Step up your social media game with powerful new capabilities, exclusively available through our new Social Media Manager Agent:

• Instagram Insights: Instantly analyze any public Instagram page, hashtags, trends, or comments accessible publicly.
  
• Google Trends Integration: Stay ahead of trends by monitoring real-time insights and popular topics.
  
• Tailored Content Creation: Effortlessly generate compelling posts, visuals, and strategies backed by data-driven analytics.
  

Enhanced User Experience
Interactive Suggestions (Improved)

• Editable Hints: Refine message hints before sending to ensure clarity and accuracy.
  
• Macro Support: Introducing the "Replace" macro, clearly highlighting potential issues and providing warnings to enhance generation quality.
  

 Live Status Updates

• Get more detailed insights into your AI's processing status.
• Enjoy delightful animated mascot icons alongside the typing indicators, enhancing interactivity and visual engagement.
  

Smart Notifications Across Platforms

Stay connected effortlessly:

• Universal Notifications: Seamlessly integrates across iOS, Android, and Desktop with instant local alerts directly in your OS or in-app as soon as chat completions are ready.
  
• One-Click Access: Notifications direct you instantly to relevant updates or messages.
  
• Background Alerts: Stay informed when your AI Agent continues processing tasks even as you navigate elsewhere.
  

Parallel Chatting (New!)

• Multi-Chat Freedom: Start as many chat sessions with your AI Agents as you want—at the same time!
  
• Live Streaming Revisit: Jump back into any ongoing conversation and pick up the live stream exactly where you left off, anytime.
  

Speed and Reliability Improvements

• Streaming Stability: Enhanced real-time AI streaming with automatic reconnection ensures uninterrupted workflows.
  
• Optimized Performance: Enhanced responsiveness, particularly noticeable on Safari and iOS.
  
  

Bug Fixes & Technical Improvements

We've addressed key technical issues to deliver a smoother experience:

• Marketplace Search Fixes: Marketplace search is now faster, more accurate, and comes with improved pagination for easier browsing.
  
• Authentication Resolved: Fixed critical Google OAuth redirect loops.
  
• Message Handling: Eliminated duplicate messages, premature stream closures, and raw SSE chunk appearances.
  
• Stream Stability: Improved network error handling for QUIC protocol errors, SSE parsing failures, and timeout management.
  UI/UX Fixes: Corrected alignment issues, notification modal persistence, and false upgrade prompts.

Version – v. 8.6

(May 6, 2025)

What’s New:

• Remote SSE MCP integration - Agents can now connect to any Model Context Protocol (MCP) server that speaks Remote SSE (e.g., Zapier, n8n, Cloudflare Workers). This instantly unlocks the full catalog of tools or workflows your server exposes.

• Rich, interactive tours – Added educational walkthroughs across the platform to help non-technical users discover features faster and feel at home.

Version - v. 8.5

(April 15, 2025)

What's New:

• Seamless Integration: Easily activate GPT Online Search by navigating to Connected Apps in your Younet profile.

• Enhanced Research: Access a broader range of online resources to gather information and insights, making your AI model more informed and effective.

• User-Friendly: The intuitive setup ensures a smooth experience, allowing you to focus on your tasks without any hassle.

Version - v. 8.4

(February 25, 2025)

What's New:

• Enhanced Onboarding Experience: We've further improved the onboarding process to help users build even better model ideas. This new flow allows for a fast and granular way of navigating the model creation process.

• Watermark Option for Paid Subscribers: Users with a paid subscription plan now have the option to disable the watermark for the chatbot, providing a cleaner and more professional appearance.

• Extended Trial Period: The trial period has been increased, allowing users to familiarize themselves with the platform at a slower pace, ensuring a more comprehensive understanding of our services.

Version - v. 8.3

(Jan 15, 2025)

What's New:

• Google Calendar and Gmail Connected Apps: Introduced connected apps for Google Calendar and Gmail, allowing users to manage their calendar and email inbox directly through the platform.
• Send Email Connected App: Added a connected app that enables models to send emails on behalf of users. This feature is particularly useful for initiating tickets via email directly from the AI widget on a website.

Bug Fixes:

• Complex Model Prompt Generation: Fixed a bug where user prompts would fail to generate for complex model instruction sets. Introduced regeneration attempts to improve reliability.
• Auth Token Expiry Handling: Resolved an issue where expired auth tokens resulted in vague error messages. Users are now redirected to the login page, and a refresh token event has been added to extend sessions, ensuring users remain authorized if they use the platform at least once a week.
• Old Chat Deletion: Fixed a bug that prevented the deletion of old chats, allowing users to manage their chat history effectively.
• TXT Parsing Special Symbols: Addressed a bug in parsing TXT files due to special symbols, ensuring accurate data processing.
•