Catalyst

Build the work. Swap the model.

Catalyst is an AI workspace where your chats, agents, workflows, mini apps, and memory are built once — then run on whatever model is best this week: OpenAI, Claude, Gemini, open-source, or your own local models. The model is a dropdown, not a dependency.

Free for personal use with Google sign-in, no credit card. Read more at voov.ai or the docs.

Launch Catalyst →

Catalyst rendering a 3D torus knot with Python

One Workspace, Every Model

Connect 12+ providers through a single interface: OpenAI, Anthropic, Google Gemini, Mistral, DeepSeek, Groq, Together, vLLM, Ollama, and any OpenAI-compatible endpoint. Native adapters for Claude, Gemini, and OpenAI mean each provider’s full capabilities — thinking blocks, vision, tool use — work the way they should.

AI Services

A Chat That Can Do Things

Switch models mid-conversation, dial reasoning effort from none to high, and run the same prompt against multiple models side by side to see which one does the job best. Chats render math, diagrams, and code, read images and PDFs, and can be shared with a read-only link.

Model Comparison

Chats Become Workflows

Figure something out in chat, then say “save this as a workflow” — the path the model discovered becomes a deterministic pipeline you can re-run for a fraction of the cost. The visual editor has ten node types (input, output, LLM, agent, Python, template, tool, condition, loop, HTTP, and sub-workflows), a JSON view, run history, and scheduling with email delivery.

Workflow Editor

Ask for an App, Get One

Describe a tool — “an app to log my food and chart it” — and Catalyst builds the workflows, writes the app, and opens it live. Mini apps run in a sealed sandbox with no internet access and only the collections and workflows you explicitly grant. Edit them by talking to the AI, import npm packages, and publish them to a public catalog where others can run or fork them.

Mini App Editor

Memory That Lasts

Catalyst remembers on purpose: a compact user profile, atomic facts recalled by meaning, and automatic session summaries. Everything is inspectable — pin what matters, archive what doesn’t. Embeddings are computed locally with LanceDB, so memory never leaves your tenant.

Memory Manager

Python Code Execution

Run Python directly in your AI conversations. Generate matplotlib visualizations, process data, and automate tasks — all executed in a sandboxed environment with full library support.

Python Code Execution

Image Generation & Editing

Create images from text with free models (z-image, Ideogram) or premium ones (Gemini, OpenAI), then keep going conversationally — edit results, change styles, or compose multiple images into one, all in the same chat.

Image Generation

Image Editing

Web Search & URL Fetching

Search the web in real time via Brave Search and fetch any URL as clean markdown — the AI chains searches and fetches together, then cites its sources.

Web Search

Ask Your Data in Plain Words

Connect PostgreSQL, MySQL, SQL Server, or Oracle — or upload SQLite and DuckDB files — and ask questions in plain language. Catalyst writes read-only SQL, returns a grid you can export to CSV or Excel, and reuses the results in chats and workflows. Drop PDFs, Word, PowerPoint, Excel, CSV, or JSON files on any prompt as knowledge.

Prompts & Skills

Save prompt templates with variables and live preview, and package the patterns that work into named skills — bundles of instructions, tools, and prompts invokable across chat, workflows, and agents.

Prompt Library

More Features