Local WebSocket bridge connecting browser extension to MCP servers (workspace folder access, file operations, terminal commands). Aggregates tools from multiple MCP servers and exposes them to AI chat sites (DeepSeek, Gemini, Kimi, GLM, Qwen, Arena, Meta AI).
OpenRender exhibits severe definition quality gaps across all dimensions. Tool names lack consistent verb-noun structure (e.g., 'list_commands', 'workspace_info' instead of 'get_workspace_info'). Descriptions are present but minimal (10-50 chars), providing insufficient context for LLM tool selection. Input schemas are visible but sparse: most tools have only 1-2 parameters with basic type info; parameter descriptions are terse. Output schemas are entirely undocumented, no indication of what fields are returned or their types. Error handling is absent from all tool definitions. The 'execute_luau' and 'run_command' tools are particularly risky: they accept arbitrary code/shell input with no validation guidance, no dry-run option, and no error recovery hints. No tool declares permissions, audit requirements, or idempotency guarantees. Composition is poor: 'list_commands' and 'list_mcp_servers' are discovery tools but lack pagination or result limits, risking context window exhaustion.
Create a new folder in the workspace
Delete a file or folder from the workspace (permanent, cannot be undone)
Edit a file by replacing specific content
Execute Luau code in Roblox Studio datamodel
Search file contents for a pattern
List all available commands with full parameter details
List contents of a directory in the workspace
List all connected MCP servers and their available commands
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., what does 'list_directory' return? file names only? sizes? timestamps?). This violates pattern:tool and forces agents to guess.
Irreversible tools (execute_luau, run_command, delete_path) lack dry-run, confirmation, or error recovery guidance. No indication of what errors are retryable vs fatal. Violates pattern:confirmation-request and pattern:recovery-guide.
Parameter descriptions are minimal (10-30 chars) and lack format/constraint guidance. E.g., 'pattern' in grep_files and search_files does not specify: is it a regex? glob? literal string? What character limits? Violates pattern:constrained-input.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | <=2025-11-25 | v2 |
Move or rename a file or folder in the workspace
Read contents of a file from the workspace
Execute a terminal command in the workspace folder
Capture screenshot from Roblox Studio (Roblox-specific tool)
Search for files matching a pattern in the workspace
Display directory tree structure
Get information about the workspace folder
Write or create a file in the workspace
Discovery tools (list_commands, list_mcp_servers) have no pagination, limit, or result-count parameters. Large result sets risk context window exhaustion. Violates pattern:paginated-result.
Tool names lack consistent verb-noun structure. 'workspace_info' should be 'get_workspace_info'; 'screen_capture' should be 'capture_screen'. Inconsistency forces LLMs to reason about naming conventions instead of inferring intent from the name alone.