Changelog

​

✨ New

• Agent Blueprint import: Import an Agent Blueprint from wonka-ai.com in one click.
• Share skills with teams: Share a skill with a whole team so everyone in your organization can reuse it.
• Spending limit per agent: Organization administrators can set a spending limit in euros for each agent.
• Longer conversation starters: Agent conversation starters can now be up to 250 characters long.

​

🔧 Improved

• Model recommendations: When you configure an agent, WonkaChat now shows which models are recommended.
• Message editor: More room to write and edit your messages.

​

🐛 Fixed

• Member search: Searching organization members now looks through the whole list, not only the page on screen.
• Organization deletion: Organization administrators can delete their organization again.
• Files in search: Files sent for search keep their settings.
• Binary attachments: Clearer message when a binary attachment cannot be read.
• Translations: English, French and Dutch texts were reviewed and duplicates removed.

​

🔒 Security and reliability

• Document viewer: Documents open more reliably in the viewer, and its configuration is now signed.

​

✨ New

• Tooltips in the collapsed sidebar: When the sidebar is collapsed, hovering an icon (New chat, Tools, Agents, History, Help, Settings) shows its name.

​

🔧 Improved

• Agents reaching their step limit: An agent that runs out of steps now ends with a summary of what it did instead of an error, and you can continue the conversation from there.

​

🐛 Fixed

• Copy message: Copying a message now keeps its images and the names of attached files, not only the text.
• Large attachments with code execution: Large files attached to a chat (usually over 10 MB, such as presentations) are now also available to agents that run code.
• SharePoint search: Fixed search failures for some agents connected to SharePoint documents.

​

🔒 Security and reliability

• Search indexing: Search indexing is more stable under heavy load.

​

✨ New

• New models: GPT-5.6 Luna and GPT-6.1 Sol are available in the model picker.

​

🔧 Improved

• Automatic model routing: The Router presets were updated. Cost stays on GPT-5.6 Luna, Balanced moves up to GPT-6.1 Sol for harder requests, and Performance moves up to Claude Sonnet 5.5. The reasoning effort is adjusted to each preset.
• Usage statistics: Usage is now recorded for every request, including those reported with no usage by the provider.

​

🔒 Security and reliability

• Billing accuracy: Costs are tracked more precisely and can no longer be counted twice.
• Stability: Better stability for long sessions.

​

✨ New

• Connected accounts: Link and manage the accounts behind your connectors from Settings, the chat, agents and schedules. The account you pick is remembered and reused.

​

🔧 Improved

• Newer models: Support for GPT-6.1 Sol and Claude Sonnet 5.5, with up-to-date context limits and pricing.
• Model selection: Automatic model selection takes better account of how complex your request is.

​

🐛 Fixed

• Memories settings: The Settings → Memories panel is back for organizations that use Memories.

​

🔒 Security and reliability

• Connector sign-in: Linking an account to a connector is more reliable, and each user’s credentials stay isolated.
• Billing accuracy: Better protection against usage being recorded or charged twice.

​

✨ New

• Notification center: In-app notifications with preferences, including a notification when someone shares an agent with you.
• Emails in agent knowledge: Agent knowledge now accepts .eml email files and reads their content.

​

🔧 Improved

• Sign-in by organization: The login page supports organization-specific sign-in, with organization selection and clearer error messages.
• Usage and cost visibility: More complete usage and cost details in organization and user statistics.
• Password recovery: Password recovery emails are always sent, whatever the organization.

​

🔒 Security and reliability

• Response streaming: Fewer interruptions while a response is being written.
• Search and uploads: More consistent search results and more stable file uploads.

​

✨ New

• Send feedback from WonkaChat: Share your feedback directly from the app.
• Navigate and quote in chat: Jump between messages, quote part of an agent’s answer, and long pasted text is added to the conversation as a file.
• PTC Creo files: WonkaChat now accepts PTC Creo models, assemblies and drawings.
• Outlook .msg files: Outlook .msg emails can now be read.

​

🔧 Improved

• Clearer tool status: While a tool runs, the status message uses plain language instead of internal service names.
• Large files: Large files are read by searching their relevant parts, so they work more reliably.
• Model prices: Model prices are based on live provider rates, so the prices shown stay current.
• Agent updates: When you confirm a change to an agent, WonkaChat understands your confirmation more reliably.

​

🐛 Fixed

• Stopped responses: Stopping a response no longer creates duplicate content.
• Connector sign-in: Each connector keeps its own sign-in redirect settings, so one connector no longer affects another.

​

🔒 Security and reliability

• Connector sign-in security: Sign-in redirects for connectors were hardened.
• Platform hardening: Several improvements to how configuration and credentials are handled.

​

✨ New

• Shared conversation links: Shared links now show a preview, can be refreshed in place, and can be public or limited to your organization.
• Humanizer skill: Humanizer is now available as a built-in skill.
• Administrator access to agents: Organization administrators can access every agent created in their organization.
• Longer agent instructions: Agent instructions can now be up to 50,000 characters long.
• Last sign-in method: The login screen highlights the method you used last time.

​

🔧 Improved

• Schedules: An agent’s schedules are paused when the agent changes, so they never run with an outdated version.
• Model retirement alerts: You are warned in advance when any model you use is about to be retired, not only chat models.

​

🐛 Fixed

• Unexpected sign-outs: Fixed an issue that regularly disconnected users.
• Faster start: Responses start as soon as the app is ready instead of waiting.
• Template chats: Template chats no longer open an outdated session.
• Share tab: The access list in the Share tab now shows up right away.
• Invitations: Pending invitations are removed when the person is already a member.
• Custom connectors: Fixed sign-in redirects for custom connectors, including organizations with their own domain.

​

🔒 Security and reliability

• Dependency updates: Open-source dependencies were updated to address known vulnerabilities.
• Unstable connections: Responses recover better when the connection is slow, interrupted or resumed.

​

✨ New

• New agent creation: A reworked flow to create, update and manage agents. It checks that the needed connectors are authorized and can create several agents at once.
• Agent templates: Groundwork for ready-to-use agent templates, translated in your language.
• Sidebar navigation: Improved navigation in the sidebar.

​

🔧 Improved

• Agent panel: A more consistent agent interface and panel.
• Connectors in agents: Tools and connectors are set up more reliably when you create or update an agent.
• Files created by code: Files produced while running code stay available for download after the session ends.

​

🐛 Fixed

• Invitations and share links: Fixed issues with sending and accepting invitations, share links and agent history.

​

🔒 Security and reliability

• Dependency update: A dependency was updated to fix a known vulnerability.

​

✨ New

• Guided onboarding: Onboarding tasks help you get started and complete the key setup steps.
• Flexible agent schedules: Schedules now support quarterly runs and custom cron expressions.
• Agent and organization management: Transfer agent ownership and manage your organization more easily.
• Manage your memories: A new Memories section in Settings lets you browse, edit and delete your personal and project memories.

​

🔧 Improved

• Connector names: Connectors show the name you gave them instead of an internal identifier.

​

🐛 Fixed

• Usage and billing: More accurate reconciliation of usage and payments.

​

🔒 Security and reliability

• Audit log for roles and permissions: Changes to roles and permissions are logged per organization.

​

✨ New

• Organization system prompt: Organization administrators can set instructions that apply across the whole workspace.
• Projects library: The projects library now works like the agents marketplace, with clearer browsing and management.
• Richer chat interactions: Clearer activity indicators and interactive cards to answer an agent’s questions and forms.
• Excel attachments: Excel files attached to a conversation are understood better.

​

🔧 Improved

• Artifact previews: Generated content is previewed more safely and consistently.
• Chat activity: Reasoning, tool calls and progress are easier to follow.
• Projects: Clearer listing, renaming and deletion of projects.
• Files and conversations: Updated controls for attachments, sharing, bookmarks and conversation management.
• Tools and models: More consistent tool labels and icons, connected apps and model selection.

​

🐛 Fixed

• Organization boundaries: Closed the remaining cases where a resource could be seen from another organization.
• Unavailable files: When a generated file cannot be delivered, WonkaChat now says so.
• Binary files: Clearer guidance when a binary file cannot be read.

​

🔒 Security and reliability

• Organization access controls: Sharing and permission checks enforce organization boundaries more strictly for agents, prompts and shared resources.

​

✨ New

• Clear message on connector access: If an agent cannot use its connectors, you are told why before the chat starts.
• More image-capable models: Additional models can now read images.

​

🔧 Improved

• Connectors stay connected: Connectors now stay connected for up to 12 hours of inactivity.
• Outlook in automatic chat: Automatic chat uses the up-to-date Outlook connector.
• Organization isolation: Agents are always tied to the organization that uses them.
• Usage limits: Usage stops as soon as a limit is reached, and billing is more accurate.
• Sign-in protection: Failed sign-in attempts are counted per user, so colleagues on the same network no longer block each other.

​

🐛 Fixed

• Chat delays: Fixed occasional delays before a response appeared.
• Connector data: Connector data is kept separate per organization.

​

🔧 Improved

• Sorting: Agents, models and skills are sorted alphabetically.
• Model pricing: An icon shows the price level of each model.
• Prompt variables: Place variables anywhere in a prompt.

​

🐛 Fixed

• Thinking loops: Agents no longer get stuck thinking in a loop.
• Image generation: Fixed image generation.

​

✨ New

• Agent list view: Browse agents as a list.
• Odoo read-only mode: Connect Odoo so agents can read data without changing anything.
• Zero-data-retention models: Models whose providers do not keep or train on your data are now available.
• New connectors: Intercom, Shopify, Meta, Miro and Google Workspace (Gmail, Calendar, Drive, Docs and Sheets in a single connector).

​

🔧 Improved

• Prompt bar: Updated message in the prompt bar.
• Odoo errors: Clearer message when the Odoo connection fails.

​

🐛 Fixed

• Large connector results: Very large results from a connector no longer overflow the model’s context.
• Double sending: A message can no longer be sent twice by accident.
• Editing during a response: Editing a prompt while a response is being written works again.

​

✨ New

• Skills: Agents can use skills to create Word, Excel and PowerPoint files, artifacts and agents. You can also create your own skills.
• Rename your organization: Change your organization’s name.
• Custom connectors with OAuth: Add custom connectors that sign in with OAuth.
• Duplicate a shared agent: Make your own copy of an agent shared with you.
• Nano Banana image generation: Generate images with Nano Banana.

​

🔧 Improved

• User management: Sort users and see countdowns in the user management page.
• Conversations: Change a previous answer to an agent’s question, better microphone interaction, and the artifact viewer closes when you switch conversation.
• Connectors: Stop a connector sign-in in progress, and the connector name is shown when you sign in.
• Display: Long agent names fit in the dashboard, pie charts use distinct colors, and recent times are shown in minutes.
• Usage: Usage resets on the first day of each month.
• Smarter agents: Code execution is only used when asked, agents suggest connecting a tool when it is needed, and Odoo tasks are assigned to the right user.

​

🐛 Fixed

• Chat: Fixed chat interruptions, conversations stuck on “New chat”, and long agent instructions being cut when pasted.
• Emails: Fixed display issues and Markdown rendering in generated emails.
• Files: Long files no longer crash, and documents reach code execution correctly.
• Agents and permissions: Fixed agent duplication rights, agent versions and user permission changes.
• Connectors and artifacts: Fixed connector editing, some connector connections and artifact generation.
• Branding and translations: The WonkaChat logo always shows correctly, and missing skill translations were added.

​

✨ New

• New generation of connectors: A new connector provider brings better tools, with a read-only option.
• Statistics dashboard: A new dashboard to follow your usage statistics.

​

🔧 Improved

• Agents: Agents are listed alphabetically, and the settings icon is hidden when you cannot edit an agent.
• Models: Models show a friendly display name.
• Questions: Go back to a previous step while answering an agent’s question.
• Artifact viewer: The artifact viewer closes when you switch conversation or tab.

​

🐛 Fixed

• Chat with an agent: Starting a chat with an agent no longer reopens the previous conversation.
• Drafts: A draft no longer stays visible after the message is sent.
• Switching agents with @: The @ no longer disappears when you switch agent.
• Interrupted chats and emails: Fixed interrupted chats and display issues in generated emails.

​

✨ New

• Code execution: Give an agent access to a secure sandbox to run code (enabled per organization).

​

🔧 Improved

• Files: Upload files to agents and prompts more easily, and upload and edit Markdown files.
• Scheduled workflows: Updated screens for scheduled workflows.
• Odoo connection window: Better layout, and your entries are kept if a connection attempt fails.
• Drafts: Each conversation keeps its own draft.

​

🐛 Fixed

• Agent sharing: Fixed an agent sharing issue.
• Files in follow-up messages: Attached files are correctly taken into account in follow-up messages.
• Questions: Agent questions no longer load forever.