Agent Desktop/Get started/Vinkius Agent Desktop
Vinkius Agent Desktop
Use local or frontier AI models, connect real capabilities, approve actions and work across your Spaces from one desktop app.
Vinkius Agent Desktop lets you work with an AI agent that can do more than answer. Choose the model you want to use, connect the services you trust, ask in plain language and watch the agent complete real work.
The app gives you one place for:
- Models: connect Ollama, OpenAI, Anthropic, Google Gemini and other supported providers, or use a compatible endpoint
- Spaces: keep conversations organized by project, with pinned chats, grouping and space statistics
- AI Capabilities: discover and connect services such as GitHub, Slack, Notion and other MCP servers
- Approval: decide whether an action is always allowed, asks first or is denied
- Agent browser: let the agent browse with a visible live preview when a task needs a website
- AI Governance: review activity, security, spend, reliability and failures
Good afternoon
AI can now review pull requests and merge changes.
Thousands of connectors put your AI to work.
Start with the way you already work
You can use a local model when you want the conversation to stay on your computer, or connect a frontier provider when you need a different model or more reasoning capacity. The app keeps model setup in one place, and you can change the model for a chat from the composer.
Give the agent useful access
Connectors provide the services your agent can use. The app shows the connector name, capabilities, credentials state and connection tokens. You decide which capabilities are enabled. The default action policy is ask, so a new capability does not silently run the first time.
See the work, not only the answer
During a response, the Canvas can show reasoning steps, connector activity, tool outputs, browser activity, strategy and memory. Select an activity to open its details in the right panel. When a task needs your decision, the app pauses and shows a confirmation card.
Your first steps
- Install and open the app
- Choose a model
- Start a chat in a Space
- Connect AI Capabilities
- Review an action before it runs
The app is built for end users. You do not need to understand MCP or write integration code to use the capabilities available in your account. For developers who want to build a connector, see MCP Fusion.