Loading ToolBench
A constrained Google Maps MCP server with bounded browser automation and optional Human takeover through a dedicated Chrome session.
Maps-browser-mcp provides 7 tools with complete input schemas and descriptions. Naming follows verb_noun convention (search_maps, get_directions, etc.). Descriptions are present and actionable (avg ~120 chars). However, output schemas are not documented in the provided source, parameter descriptions lack detail on constraints/formats, and error handling guidance is minimal. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Tools are well-composed (single responsibility) and include natural identifiers (addresses, place names). Missing: pagination guidance, field-level output documentation, and recovery hints in error scenarios.
Get directions between two locations on Google Maps with support for different travel modes
Get detailed information about a specific place on Google Maps
Read the current visible state of the Google Maps interface including visible places, markers, and UI elements
Save a place to a list in Google Maps (requires user authentication and approval)
Search for places, addresses, or points of interest on Google Maps
Send a route to a device or share it (requires user authentication and approval)
Display a map view centered on a specific location or showing search results
Output schemas not documented. Tools return data but LLMs cannot plan downstream calls or extract fields without trial-and-error. E.g., search_maps returns what fields? Does it include place_id for chaining to get_place_details?
Parameter descriptions lack format/constraint details. E.g., 'zoom' accepts 1-21 but description omits range. 'fields' array in get_place_details lists examples ('name, address, phone...') but no enum or validation guidance. LLMs will guess invalid values.
No tool annotations despite explicit risk classifications. save_place and send_route are marked WRITE but lack destructiveHint annotation. LLMs cannot distinguish safe (read-only) from risky (write) tools without annotations.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Error handling lacks recovery guidance. No indication of what to do if search_maps returns no results, or if get_directions fails due to unreachable destination. Errors should guide LLM to alternative tools or user input.
Pagination not addressed. If search_maps returns 100+ results, no limit or cursor guidance provided. Large result sets blow context windows and degrade LLM reasoning.