Visual feedback for your AI coding agent

Connect Claude Code, Codex or Cursor to Vitrinka. Publish screens, receive precise feedback and return proof of the changes.

Start free

Connect the tools you use

Install the Vitrinka CLI and connect your supported MCP client. Every client connects to the same address, your deployment's /mcp, and signs in with its own OAuth flow. That covers Claude Code, Codex, Cursor, OpenCode, VS Code, Gemini CLI and hosted assistants. vitrinka setup writes the entries, and nothing runs locally. Your team's screenshots and feedback then have a shared destination outside the chat transcript.

Read the guide

Keep feedback connected to the screen

Publish screenshots and artifacts to a board. A teammate can point at the result and leave a focused annotation. During live pairing, an attached agent receives work from the board and can respond with the changed screen. People and agents share the same evidence rather than maintaining separate lists of what needs fixing.

Read the guide

Make the workflow repeatable

Use the CLI for published sets and the skills for capture, review and recorded-session workflows. The server prepares the information agents consume; the references describe the current commands and tools. Pick a task up in the repository and hand it back when you are done; the task keeps the branch, the runs and the proof without a status update. Start with one board in one repository, then carry the same workflow into the team's regular development and review process.

Read the guide Your day, in depth