Features
What the workspace ships with.
No feature-list theater. These are the surfaces in the app today, described the way you'd meet them.
Multi-provider chat
Talk to OpenAI, Anthropic, Google Gemini, Ollama for local models, DeepSeek, Groq, OpenRouter, xAI, Mistral — or any custom OpenAI-compatible endpoint. Model lists are fetched per provider where the API supports it.
An agent that acts on your checkout
The built-in tool catalog covers a terminal, file tools (read, edit, search, glob, grep), git tools (status, diff, commit, apply patch), GitHub pull requests and issues, typecheck / lint / test runners, browser automation, notebook editing, and language-server queries.
Agent instances & worktrees
The agent can fan work out to child instances. Every child runs on its own git worktree branch and the result merges back into the parent branch — parallel work without clobbering your checkout.
Local Whisper dictation
Voice dictation runs entirely on your machine — Whisper via transformers.js on the onnxruntime-node backend, in the Electron main process. No audio ever leaves the app.
Skills & marketplace
Skills ship with the app as marketplace resources and can be loaded into a run. Plugin rules are supported alongside skill files.
MCP client
Connect Model Context Protocol servers, browse their tools, resources, and prompts, and pin the tools you want into the agent's catalog.
Self-built agent tools
When no existing tool fits, the agent writes its own at runtime. A new tool becomes available immediately in the catalog — no restart — and runs in an isolated child process. Built tools show up under Settings → Agent tools.
Long-term workspace memory
The agent keeps structured notes under .vyotiq/memory/ in your workspace and re-reads them on later runs — durable facts survive across sessions.