Agent guide / Cursor

Turn what you circle into context Cursor can use.

Capture the exact UI state, keep your narration aligned to each mark, and hand Cursor a report made of discrete, actionable findings.

Path one: paste a report into Cursor

Record while you test, then paste the copied Markdown path into Cursor. Cursor reads the report and the colocated evidence images without needing you to describe the layout twice.

Pick the exact window

Press Cmd+Shift+F, choose Window, and click the app you are reviewing. MarkuprPlus remains locked to that window as it moves or resizes.

Talk and mark

Narrate what is wrong. Hold Cmd and drag to draw a freehand, circle, or highlight mark; release it to create one finding for that moment.

Stop and paste the path

Press Cmd+Shift+F again. The Markdown report is written locally and its path is copied to the clipboard, ready to paste into your agent.

Path two: expose capture tools over MCP

Configure the local MCP process with the same package command in any Cursor MCP configuration surface that accepts a command and argument list.

capture_screenshot
capture_with_voice
describe_screen
start_recording
stop_recording
analyze_video
analyze_screenshot
push_to_github
push_to_linear

MCP server process

npx --yes --package markuprplus markuprplus-mcp

Path three: choose how the report is structured

Cursor Agent CLI runs as a provider in non-interactive, read-only Ask mode. It receives report material and cannot turn the analysis pass into an editing session.

The current app exposes the following implemented providers. Local Rules remains the zero-setup fallback if the selected model or CLI is not reachable.

Codex CLI
Claude Code CLI
OpenCode
Cursor Agent CLI
Qwen Code
Goose
Amp
Kiro CLI
Aider
Ollama
LM Studio
Anthropic API
Local Rules