Loading ToolBench
A relay-hub MCP server that enables remote control of desktop agents via WebSocket, supporting shell execution, screenshots, filesystem operations, process management, and system information retrieval
rc-mcp provides 17 tools with complete input schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention well (shell_exec, fs_read, process_list). Descriptions are present but often generic, most lack context on WHEN to use the tool or what happens next. Parameter descriptions are minimal (e.g., 'Process ID to query' for process_info.pid). Output schemas are not documented in the visible code. Error handling guidance is absent, no recovery hints or categorization. Security-critical tools (shell_exec, fs_delete, input_keyboard) lack confirmation/dry-run patterns. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels. Composition is reasonable, tools are single-purpose, but chaining IDs are not always returned (e.g., process_list returns PIDs but no parent process context for hierarchical operations).
Delete a file or directory on a connected agent's filesystem
List contents of a directory on a connected agent's filesystem
Read the contents of a file from a connected agent's filesystem
Get file or directory metadata from a connected agent's filesystem
Write contents to a file on a connected agent's filesystem
Send keyboard input (key presses, text) to a connected agent's display
Send mouse input (movement, clicks) to a connected agent's display
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields without knowing response structure.
Destructive/irreversible tools (shell_exec, fs_delete, process_signal, input_keyboard) lack confirmation or dry-run patterns. No tool annotations (destructiveHint) to warn LLMs.
Parameter descriptions are minimal and lack format/constraint guidance. E.g., 'Process ID to query' does not explain valid range or error cases. No enum constraints for string parameters (e.g., signal names, image formats).
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2025-06-18+ | v2 |
Get detailed information about a specific process on a connected agent
List all running processes on a connected agent with details like PID, name, and resource usage
Send a signal to a process on a connected agent (e.g., SIGTERM, SIGKILL)
Capture a screenshot from a connected agent's display
Subscribe to continuous screenshot updates from a connected agent's display
Execute shell commands on a connected agent's desktop session with optional confirmation and audit logging
Close an active shell session on a connected agent
Start an interactive shell session on a connected agent
Write input to an active shell session on a connected agent
Retrieve system information from a connected agent including OS, architecture, hostname, and resource metrics
Tool descriptions lack context on WHEN to use each tool and what happens next. E.g., shell_exec says 'Execute shell commands' but does not explain timeout behavior, output capture, or error recovery.
No error handling guidance. Tools do not document what errors are retryable, user-fixable, or fatal. No recovery hints (e.g., 'If file not found, try fs_list() to discover available paths').