Changelog

Everything Viktor learns to do, the day it ships.

Oct 31Links open right beside your chat Subscribe

June

4releases

Jun 12Fri · #029

Agent smarts Admin & security Slack, Teams & email

Ultra preset, enterprise spend limits and @mention when done

A new Ultra quality preset in the model picker, monthly spend limits for enterprise admins, and Viktor now @mentions you when your task is done.

Ultra preset. A new quality preset in the model picker that routes your tasks to the most capable models when you need maximum reasoning depth.
Enterprise spend limits. Enterprise admins can now set monthly spend limits for their workspace to keep usage predictable.
@mention when done. Viktor now @mentions you when it finishes a task, so you get notified the moment your result is ready.

Jun 10Wed · #028

Agent smarts Integrations

Claude Fable 5 and the Supabase integration

Claude Fable 5 is generally available, and a one-click Supabase integration opens up your database, edge functions and project settings to Viktor.

Claude Fable 5. Anthropic's most powerful model graduates from early access to general availability. Select it in the model picker when you need peak reasoning. It runs at 2× Opus pricing and includes automatic Vertex provider fallback for extra reliability under load.
Supabase integration. Connect your Supabase project with one-click OAuth and give Viktor 20+ native tools covering table CRUD, the Management API, migrations, edge functions, log inspection, and TypeScript type generation, no API keys to paste, no project refs to look up.

Jun 3Wed · #027

Agent smarts

Workspace personalization settings

Set a personality and custom instructions that shape how Viktor responds across your entire workspace.

Workspace personalization. A new section in Settings lets you pick a personality preset and write custom instructions that apply to every Viktor conversation in your workspace. Use it to adjust tone, set standing context ("we're a B2B SaaS company", "always reply in Spanish"), or reduce sycophancy. The instructions are injected into Viktor's system prompt so they take effect immediately, no restart needed.

May

1release

May 30Sat · #026

Agent smarts

Claude Opus 4.8

Anthropic's latest frontier model is now available in Viktor's model picker.

Claude Opus 4.8. Anthropic's newest flagship model is now available in the model picker for all workspaces. Select it when you need maximum reasoning depth. It uses adaptive thinking that scales compute to match problem complexity.

April

5releases

Apr 29Wed · #022

Agent smarts Billing

GPT-5.5, bigger Claude outputs and a refreshed Usage page

GPT-5.5 joins the picker, Claude writes much longer outputs, signup shows what you're installing, and Usage gets a Scheduled Tasks tab.

GPT-5.5 in the model picker. OpenAI's newest model is now available for selection across coworker tasks. Kimi K2.6 and Gemini Flash 3 have been retired from the default-model picker.
Longer Claude outputs. Output caps for Claude Opus 4.6 / 4.7 and the Sonnet 4 family were raised significantly, so Viktor can now return much longer reports, drafts, and analyses in a single response without truncation.
"You're installing X" on signup. When a logged-out user clicks Install on a skill or bundle, the signup page now shows exactly what's being installed, so the install intent survives the sign-in handoff and lands you straight into the right setup.
Refreshed Usage page. The Usage settings now include a Scheduled Tasks tab listing every running scheduled task in one place, plus per-user usage stats so workspace admins can see exactly which teammates are spending credits and where.

Apr 23Thu · #019

Agent smarts

Per-model reasoning controls and GPT Image 2

Pick reasoning effort per model, choose defaults from a redesigned grid, and generate sharper images with OpenAI's newest model.

Per-model reasoning controls. The Default AI model settings page is now a redesigned card grid where you can pick a reasoning effort (Minimal, Low, Medium, or High) for each model, so you can tune depth vs. speed for how you actually work.
GPT-Image-2 as the default image model. Image generation now uses OpenAI's newest gpt-image-2 by default, with better fidelity and more accurate token-based cost accounting.

Apr 22Wed · #018

Integrations Agent smarts

Reconnect integrations, smarter approvals and Kimi K2.6

One-click reconnect for broken integrations, always-approve for trusted tools, a softer past-due experience, and a new Moonshot model.

One-click reconnect for broken integrations. The integrations page now flags broken OAuth connections with a clear status and lets you reconnect in place. Your row-scoped settings are preserved, so you don't have to reconfigure Meta Ads, Google Ads, Stripe, Linear, Granola and others from scratch.
"Always approve this tool" in Slack. When Viktor asks for permission to use a tool in Slack, you can now choose to always approve future uses of the same tool. The preference is saved to your MCP settings so you stop getting re-prompted for tools you trust.
Atlassian pre-connect setup guide. Connecting Jira or Confluence now walks you through the steps your workspace admin needs to take first (whitelisting the Rovo MCP domain) before the OAuth flow kicks in, so the connection works on the first try.
Moonshot Kimi K2.6 available. Moonshot's newest Kimi K2.6 model is now selectable in the model picker (marked "in testing") alongside the existing frontier models.
Clearer account deletion for admins. When you're the last admin deleting a workspace, Viktor now shows exactly which teammates will lose access and how many scheduled tasks each person has, so you know the impact before you confirm.
Softer past-due experience. If a payment fails, you'll now see a banner at the top of the app instead of a full lockout. You can keep browsing your usage and settings while the payment recovers.

Apr 21Tue · #017

Agent smarts Integrations

Claude Opus 4.7 and an instant integrations page

Anthropic's newest frontier model is live for everyone, and the integrations page now loads instantly.

Claude Opus 4.7 available. Anthropic's newest frontier model is now available in Viktor for everyone, with a 1M context window, adaptive thinking, and summarized reasoning output.
Instant integrations page. Your integrations list now renders immediately when you open or reload the page, then refreshes in the background, no more waiting for it to load each time.

Apr 8Wed · #016

Agent smarts

Curated skill bundles

Discover curated groups of prebuilt skills in the Skills Directory.

Skill bundles. Browse curated groups of related prebuilt skills in the Skills Directory, with bundle detail pages that make it easier to understand what’s included and find the right skills faster.

March

4releases

Mar 20Fri · #014

Agent smarts

Speech and video generation

Viktor can now generate spoken audio, transcribe media files, and create AI-powered videos.

Text-to-speech & audio transcription. Generate spoken audio from text and transcribe audio or video files directly in the chat. No API keys or setup needed.
Text-to-video generation. Create videos from text prompts, powered by xAI Grok, Gemini Veo 3.1, and OpenAI Sora 2 Pro.