# Latest updates **April 16** ### :rocketchat: Interactive Chat nexos.ai can now pause during a conversation to ask follow-up questions when more context is needed, helping you get **stronger results without having to provide every detail upfront**. This means you can: * get more relevant responses when your request is broad or missing context * let nexos.ai clarify details like audience, format, time period, tools, or output type before continuing * reduce repeated prompting and manual back-and-forth * reach the expected outcome faster Available in [Chat](https://workspace.nexos.ai/chat). *** **March 12** ### :robot: Agent Builder and Scheduling You can now create an AI Agent by simply describing what you need in plain language. Then set a schedule and let it run automatically in the background. This means you can: * Create Agents without complex setup or forms * Schedule recurring work (hourly, daily, monthly, or yearly) so you stop triggering agents manually * Track every run in the Activity Log, so you always know what happened You can find it in [Agents](https://workspace.nexos.ai/agents/new) (create a new agent from scratch, then add a schedule in the agent settings). *** **March 6** ### :coins: User Memory and Personalization nexos.ai can now remember helpful context about you across Chats and Agents, so you spend less time repeating yourself and get more consistent answers.\ \ This means you can: * Keep useful context between conversations with Memory * Add your role and preferences so answers fit your style * Use custom instructions (write your own or choose suggested options) to guide how responses are written * Stay in control by turning Memory on or off and managing what is saved You can find it in Settings → [Personalization](https://workspace.nexos.ai/user/settings?tab=personalization). *** **February 26** ### :coins: API Credits Add-on You can now purchase API credits as an add on for nexos.ai. API usage is prepaid and runs on credits, where 1 euro equals 1 credit. The API works while you have credits available, and you can always see your remaining balance in Billing. API access requires a Pro plan, and Pro users will only have API access after they purchase credits. *** **February 17** ### :magnifying-glass-arrows-rotate: Deep Research Users can now use Deep Research to automate complex web searches and generate comprehensive, cited reports in minutes. The feature plans a search strategy, browses multiple sources across the web, and synthesizes everything into a structured report directly in your chat interface, saving you hours of manual data gathering. *** **February 6** ### :file-half-dashed: AI Agents Users can now create AI Agents from scratch or start fast with pre-made templates to automate routine tasks. Agents can then be connected to your tools like Google Drive, Gmail, SharePoint, Slack, and more. Key capabilities: * Create a no-code AI Agent in minutes, with a simple setup any team can master (Marketing, Sales, HR, etc.) * Use ready-made templates or build from scratch * Connect knowledge sources and integrations for real-time context * Share AI Agents with your team * Choose from top AI models and switch anytime *** **January 30** ### :wand-magic-sparkles: Smart Routing (Auto AI Model Selection) Smart Routing is now available in Chat. Just pick "Auto" from the model menu and the system will automatically choose the most suitable organization-approved AI model for each message, based on your prompt and available models. You can switch back to a specific AI model anytime by reopening the model menu and choosing a different model. *** **January 29** ### :copy: File Generation & Analysis Code Interpreter is available in Chat, Projects, and Agents. It helps users to analyze data, create charts, and generate downloadable files right inside nexos.ai. Common use cases include cleaning and exporting CSVs, summarizing trends, creating reports, and finding data errors. *** **December 18** ### :sidebar-flip: **Browser extension** Browser extension for nexos.ai released for all Chromium-based browsers (Chrome, Edge, Brave, etc.) Key capabilities: * Page summarization with a single click * Chat directly about webpage content * Access to the same models available in nexos.ai Workspace * Browser chat history syncs with nexos.ai Workspace [Get it from Chrome Web Store](https://chromewebstore.google.com/detail/nexosai-make-ai-work-wher/oekflhblfcpkcopglgkomdcaopmejfdm) *** **December 11** ### :mobile: **Mobile apps** Mobile apps for nexos.ai have been released for iOS and Android Features include: * Access to all models and assistants available in the web app * Synced conversations between web and mobile * Photo and file analysis capabilities [Download on iOS](https://apps.apple.com/us/app/nexos-ai/id6751948632) [Download on Android](https://play.google.com/store/apps/details?id=ai.nexos.app\&pcampaignid=web_share) *** **November 27** ### :chart-tree-map: **Integrations** New integrations are now available for Microsoft SharePoint, Atlassian (Confluence, Jira, Compass), Slack, and HiBob with more coming soon. Integrations are currently supported in Chat, and company owners can enable them under “Management / Integrations” in order for employees to use them. All Integrations inherit the original work tool permissions and accesses - an employee will only be able to access the files they have permissions to view/edit. *** **November 10** ### :shield-check: **Company-wide guardrails** Set organization-wide guardrails once, override per model, or exclude models entirely - with a clear priority: Excluded → Custom → Company. Configuration and visibility are now centralized in the Guardrails tab and Management → Models/Teams, replacing the old per‑model toggle. *** **October 30** ### :sparkles: **Auto-generated instructions in Assistants** Assistant setup just got easier with auto-generated instructions. Simply describe what you want your assistant to do in plain language, and we'll automatically craft detailed instructions for you. No more staring at blank fields or guessing what format to use - just explain your goal, review the generated instructions, and customize as needed. *** **October 17** ### :airplay: **Streaming in Model Comparison** Model Comparison now streams responses in real-time. Watch as each model generates its answer character by character, making it immediately clear which models respond faster and how they construct their responses differently. The new streaming interface also makes the comparison experience more dynamic and engaging. *** **October 17** ### :user-magnifying-glass: **Full list of Organization users** User management now includes a unified view across all teams. Instead of jumping between teams to manage individual users, you can now access and manage everyone from a single, comprehensive list. Search, filter, and update user permissions across your entire organization in one place. *** **October 17** ### :screwdriver-wrench: **Log time selection** + **Log spans** We've expanded observability with Log Spans and extended visibility windows. Log Spans now provide line-by-line performance breakdowns for every request, giving you granular insights into exactly how your queries are processed. Plus, we've extended data retention significantly—logs are now available for 7 days (up from 1) and metrics for 90 days (up from 30). *** **October 10** ### :shield: **Guardrails for free trial companies** Free Trial companies can now access and configure Guardrails. This update allows trial users to create, test, and apply validators (such as Regex and No PII) to control model behavior and protect sensitive data. *** **October 3** ### :user: **Assistants in compare models** We have added support for using custom assistants in Compare Models. You can now test assistants side‑by‑side with other models to evaluate performance and output differences. *** **October 3** ### :folder-image: **Image gen models** We have added support for new image generation models: gpt-image-1 and imagen 4.0. These models are now available in the model selection menu for generating high‑quality AI‑created images.
*** **September 16** ### :face-thinking: **Reasoning** Reasoning is now available across **chat**, **projects**, and **assistants**. You can now turn on a reasoning toggle to see when the model uses its internal “thinking” loop. This is how the model breaks down your query, decides which tools to use, and pulls everything together into a stronger response. The result: more accurate, higher-quality answers from any model.
*** **September 16** ### :folder-image: Enhanced image recognition We've improved our image data parsing in projects. Models can now better understand what's in pictures, identify image content, and extract text from handwritten notes and scanned documents for newly uploaded images, making visual information in your projects searchable and actionable. *** **September 3** ### :message-check: **A**nnotations & sources We've improved annotation parsing and display: annotations now process more accurately and show multiple sources with clear citations, making it easier to track where information comes from. Available in Chats and Projects.
*** **August 29** ### :paintbrush-pencil: Creativity controls in Projects Not every task needs the same approach. Some demand strict facts, others thrive on fresh ideas. Now you can set the right creativity level for each Project: * **Grounded** – facts only * **Guided** – factual with light flexibility * **Balanced** – blend of accuracy and creativity * **Analytical** – structured, data-driven responses * **Creative** – open, idea-driven exploration Pick the mode that matches your work. More control, better outcomes. *** **August 26** ### :share: **Assistants & Projects sharing** Assistants can now be shared with teams or individual users.\ \ Choose whether others can **view or edit** your Assistants and Projects, with role-based permissions for better control. *** **August 12** ### :sparkles: **Assistants now available** You can now create **reusable custom assistants with their own role, rules and model.** Add your files and knowledge → set instructions → get an AI Assistant built for your tasks. [**More on how-to in our docs →** ](/workspace/agents.md) *** **August 8** ### :screwdriver-wrench: **Improvements** Improvements on Project files handling - less errors, quicker processing. Better search results provision. Better markdown parsing. *** **August 6** ### :sparkle: **New models added** *Claude `Opus 4.1` and* `Llama 4` *is now available.*\ Enable it in **Management → Models** to start generating with the newest Claude engine. *`gpt-oss-120b` and `gpt-oss-20b` are now available.*\ Find them under **Management → Models** to enable and start using. \ Free-trial workspaces already have both `OSS` models switched on. *** #### July 31 ### :folder-open: Projects now available You can now create Projects: dedicated workspaces where AI works with your files, your context, and your knowledge. Upload documents, start a conversation, switch models and everything stays in one place.\ Your files don’t mix across projects. Your history doesn’t reset.\ **Each Project keeps its own memory, so you can pick up where you left off, with answers grounded in the data you’ve provided.** --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.nexos.ai/changelog/changelog/latest-updates.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.