zen_mcp · job 70%shortcuts_mcp · job 82%claude_command_runner · job 77%zmcp_tools · job 76%desktop_pilot_mcp · job 74%webclaw_2 · job 73%gdd · job 72%plonk_2 · job 74%MCP DeskManager · job 77%Claude Background Control · job 81%Computer MCP · job 74%native-cua · job 73%Claude Voice Bridge · job 80%Auten · job 75%iClaude · job 73%agentfenster · job 79%swift_mcp_gui · job 77%desktop-hub · job 75%PcGöz · job 74%mcp_browser_agent · job 79%open_claude_in_chrome · job 78%desktop_commander_mcp · job 74%cybersec_toolkit · job 72%chrome_devtools_mcp_2 · job 74%applescript_mcp · job 70%airmcp · job 73%computer_use_mcp · job 71%Phosphor · job 79%uitap · job 76%remote-tools · job 77%Sloth MCP · job 83%Deixion · job 80%Claude Codex Computer Use · job 78%Warp Command Runner · job 75%Chrome Controller for Claude MCP · job 75%SRC Hunter · job 75%Machines as Tools · job 74%OS Control · job 75%MainStage MCP · job 74%Claude Browser · job 72%Waldo · job 72%shellway · job 77%macOS MCP Plugin · job 79%Percentages are graph-similarity metadata, not suitability scores. A graph edge answers “is it connected” — plan fit and comparable cost are checked below.
Pick the exact reference and alternative series. Each history is separate; gaps stay gaps.
This view calls the same get_evidence tool your agent would — chart, table and MCP response describe the same records. Watch the reference series