Connect your own OpenRouter account for direct access to 300+ AI models, plus faster approvals and cleaner Space URLs.
Native OpenRouter integration. Connect your own OpenRouter API key and Viktor can chat with any of OpenRouter's 300+ models, billed to your own OpenRouter credits, with a model browser and key-status check built in.
Grouped approval cards in app chat. A burst of tool calls needing approval now folds into a single card instead of a stack, with an "Always approve this tool" option, bringing app chat's approval flow to parity with Slack.
Readable Space URLs. New Viktor Spaces get a human-readable address (e.g. workstreams-zeta.viktor.space) instead of a random string, making them easier to share and recognize.
Ask Viktor for a transparent PNG or WebP and GPT Image 2 will generate cutouts, logos, and stickers with real alpha, no manual background removal.
Transparent image backgrounds. Viktor can now generate images with a transparent background using GPT Image 2, its default image model. Just ask for a cutout, logo, or sticker as a transparent PNG or WebP and Viktor requests it directly from the model instead of removing the background afterward. Existing image requests are unaffected, transparency is opt-in.
Serve Spaces and published pages on your own domain, and generate 30-second video clips with Seedance 2.5.
Custom domains for Spaces. Any workspace admin can now point a domain they own at a Space from its Settings tab, no waitlist or flag. Viktor walks you through the DNS records, provisions the certificate, and keeps whatever access policy the Space already had. Published pages get the same treatment: ask Viktor to add a domain for your pages and one hostname serves your whole viktor.page namespace, with per-page access carried over.
Seedance 2.5 for video generation. The default text-to-video model is now Seedance 2.5: single-shot clips up to 30 seconds at 720p, and up to 30 reference images to steer the look. A cheaper 480p variant is available for drafts, and Seedance 2.0 stays selectable when you want the lower price.
Upload, edit and download shared skills from the Skills page, run any of them with / in chat, and use the API from Microsoft Teams.
Upload, edit and download shared skills. The Shared Skills tab on the Skills page now does more than show you what Viktor knows. Anyone in the workspace can download a skill as a zip with its full files, admins can edit any text file in place (including SKILL.md) and a new Upload skill flow takes a folder or zip and validates it file by file before it lands. Edits take effect on Viktor's next turn.
Shared skills as slash commands. Typing / in the chat composer now lists your workspace's shared skills alongside system commands, so picking one sends it straight to Viktor. Autocomplete follows the / wherever it sits in your draft, and recognized commands get highlighted as you type and in the sent message.
Microsoft Teams on the API. The OpenAI-compatible and public APIs now work for Teams workspaces exactly as they do for Slack: chat completions, responses, messages, threads and runs all resolve your Teams identity and run with your connections. No setup or opt-in needed.
Moonshot's Kimi K3 is now available in the model picker, and Viktor's OpenAI-compatible API is generally available for all workspaces.
Kimi K3. Moonshot's 2.8-trillion-parameter reasoning model joins the model picker with 1M-token context and always-on chain-of-thought. Choose between low, high, and max reasoning depths depending on the task. It runs at roughly half the cost of comparable frontier models.
Public API. Viktor's OpenAI-compatible API is now generally available. Call it using the OpenAI SDK, Anthropic SDK, or any compatible client to run Viktor programmatically. Same models, tools, and integrations as the chat, accessible with a personal API key from Settings.
Anthropic's most powerful model is now available in Viktor's model picker.
Claude Opus 5. Anthropic's latest flagship model, launched July 23, is now available in the model picker for all workspaces. Select it when you need the deepest reasoning and most capable generation Anthropic offers. It runs on the direct Claude API with a 1 M-token context window.
OpenAI's newest model family is now available in the model picker, three models spanning peak reasoning to fast, affordable runs.
GPT-5.6 Sol, Terra & Luna. OpenAI's latest model family is now available in the model picker for all workspaces. Sol sits at the top of the lineup for maximum reasoning depth, Terra offers a strong balance of capability and cost, and Luna is the fastest and most affordable of the three. All three are marked Beta while OpenAI finalises the release.
Anthropic's Claude Sonnet 5 is now available in the model picker for all workspaces.
Claude Sonnet 5. Anthropic's newest Sonnet-class model joins the model picker for all workspaces. Pick it for a strong balance of capability, speed, and cost.
After installing a skill or bundle, a new Connect tools checklist walks you through the integrations it needs, without blocking you from starting.
Guided tool setup. Installing a skill or bundle now shows a non-blocking "Connect tools" checklist of the integrations that skill needs. Connect them right away or skip ahead and come back later. Nothing blocks you from getting started.
