Lovision MCP Overview

Lovision MCP connects AI clients like Codex, Claude, Cursor, and VS Code Copilot to your Lovision projects through the Model Context Protocol. Once connected, an agent can read project and canvas context, create or edit designs, check quality, capture previews, and export results with your authorization.

MCP is not specific to one AI client. Any client that supports remote Streamable HTTP MCP can use the same Lovision MCP URL and tools.

Early Access

Lovision MCP is currently in Early Access, not a public self-serve feature. Self-serve OAuth sign-in is not fully available yet; if MCP has not been enabled for your account or client, adding the server may not show a Lovision / OneAuth authorization flow. Before starting, confirm that Lovision has provided access and an authorization method.

How It Works

Lovision MCP is session-first:

1. The AI client connects to https://mcp.lovision.ai/.
2. You authorize the connection through Lovision / OneAuth or the Early Access authorization method Lovision provides.
3. The agent calls lovision.session.init.
4. The agent calls lovision.capabilities.list to inspect what tools are available.
5. The agent calls lovision.document.getContext to read the current project, page, selection, viewport, fonts, tokens, and capability summary.
6. The agent performs your task, such as creating a design, editing a page, exporting an image, or managing a project.
7. For meaningful design edits, the agent should run lovision.quality.check and lovision.preview.capture before summarizing the result.

What The Agent Sees

Lovision MCP gives the agent structured design context, not just screenshots. Context can include:

• Current project, page, and document target.
• Current selection, node summaries, and viewport information.
• Available fonts, components, design tokens, and page background.
• Tools, permissions, limitations, and suggested next actions for the current session.
• Quality findings, preview screenshots, and export artifacts.

This helps the agent understand what the design is and what it is allowed to do, rather than guessing from pixels alone.

Common Workflows

Create Or Edit Designs

You can ask an agent to create new design sections, adjust an existing page, make campaign graphics, pricing cards, social assets, presentation pages, or brand visuals. For complex layouts, the agent can validate a Lovision Blueprint before applying it to the canvas.

Read And Summarize Projects

The agent can list projects, open a project, read the current page context, and summarize page structure, design tokens, fonts, and potential quality issues.

Check Quality And Preview

After writing to a design, the agent can check layout, text, media, color, and renderability issues, then capture a preview screenshot so you can review the result.

Export And Deliver

Within the permissions you grant, the agent can export images or other supported design artifacts and finish with a concise work summary.

Manage Projects And Collaboration

When authorized, the agent can help list projects, create projects, move folders, inspect sharing state, invite collaborators, or update collaborator roles.

Tool Groups

See Lovision MCP Tools for the full list.

Get Better Results

• Ask the agent to read context and summarize the target before changing a design.
• For complex tasks, ask for a short plan and Blueprint validation before applying changes.
• After edits, ask the agent to run quality checks and capture a preview.
• Confirm high-impact actions such as export, sharing, deleting, or inviting collaborators.
• If the agent is unsure what is available, ask it to call lovision.capabilities.list.

Next Steps

• Connect Lovision MCP — Connect your AI client to Lovision.
• Lovision MCP Tools — See what each tool group can do.
• Agent Skill — Install the skill that teaches your agent the correct Lovision MCP workflow.