mcp-see-screen
Related Servers
Alternatives to mcp-see-screen
No user-submitted related servers found.
Related Servers
- FlicenseAqualityDmaintenanceEnables LLMs to capture screenshots and screen recordings through MCP with chunked session-based transfers for reliable image consumption. Supports multi-monitor selection, timeline capture, and compatibility with both vision and non-vision language models.111-
- AlicenseNot gradedqualityCmaintenanceEnables MCP-capable agents to observe a controlled device's screen from captured video and operate it by sending simulated keyboard, text, scroll, drag, and mouse input. Screenshots can be inspected as numbered visual targets, with dense areas split into zoomed regions so the agent picks a target ID and the server handles coordinate mapping and click execution.1MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to capture, analyze, compare, and share authorized screen content using multimodal models, with local consent, sensitive-data masking, and structured MCP resources for cross-agent reporting.Apache 2.0
- AlicenseAqualityBmaintenanceEnables AI agents to perceive and quantitatively verify physical hardware through USB webcams and RTSP/HTTP network streams, returning inline image content plus numerical metrics such as brightness, contrast, edge density, motion, SSIM baseline comparisons, and OCR text. Agents can watch feeds for visual conditions, correlate buffered frames with serial/log/MQTT events, and apply privacy masks and access controls.1966 PyPIMIT
- AlicenseAqualityBmaintenanceMCP server that captures short screen recordings as numbered, timestamped still frames and composes them into a single contact sheet returned as image content, letting AI agents reason about motion, animations, and UI transitions instead of inspecting video frame by frame. Also supports single screenshots and listing targetable windows.653 npm1MIT
- AlicenseNot gradedqualityDmaintenanceLets an LLM see what's on your screen by capturing an entire monitor or a specific application window.MIT
TDQS
Scored across 5 tools
Each tool has a clearly distinct role: see_screen captures, see_region restricts capture, see_stop halts it, see_status reports runtime diagnostics, and see_help is declarative documentation. The only mild adjacency is see_status vs see_help, but the descriptions explicitly separate 'can't see the screen' troubleshooting from parameter-selection guidance.
All five tools use the same see_ + noun pattern (see_screen, see_status, see_region, see_stop, see_help), giving a predictable, readable surface with no mixing of conventions.
Five tools map exactly onto the lifecycle needs of a screen-capture server: capture, scope, halt, inspect state, and read docs. Nothing is redundant and nothing feels padded.
The surface covers capture (watch/peek), region scoping, stopping, status, and documentation, which handles the core workflow end to end. Minor gaps remain, such as no tool for persisting default quality/pace settings or managing capture source selection outside of per-call parameters.