Loading ToolBench
Model Context Protocol server exposing AtlasTrip travel inventory as tools. Read-only access to flight and lodging inventory, employee directory and cost centre budgets.
AtlasTrip's travel inventory MCP exposes 7 well-named, read-only tools with complete input schemas and solid descriptions. Tool names follow verb_noun convention (search_flights, list_airports, lookup_employee). All 7 tools have descriptions (avg ~120 chars, within 10-1024 baseline). Input parameters are typed and described. However, output schemas are not documented in the source, the server returns data but LLMs cannot see what fields to expect. Error handling is absent: no recovery guidance, no categorization of retryable vs fatal errors. Parameter descriptions lack format constraints (e.g., 'YYYY-MM-DD' is stated but not formalized as a pattern). No tool annotations (readOnlyHint, idempotentHint) despite all tools being read-only. Missing pagination guidance for list_airports and search results. Overall solid foundation but lacks production-grade error handling and output documentation.
Current budget position of a cost centre: the quarterly allowance, everything already committed or spent against it, and what is left.
Ground transport options between the airport and the city centre.
List the airports AtlasTrip has inventory for.
Look up an employee by email. Returns their grade, home airport, passport country, cost centre and manager, which the other agents need in order to apply policy and route approvals.
Look up a customer site by its venue id.
Search bookable flights on one city pair for one departure date. Returns the cheapest fare products first, with the total fare (base plus taxes), cabin, duration, stop count and modelled CO2.
Output schemas not documented. LLMs cannot see what fields search_flights, search_hotels, or lookup_employee return, forcing them to guess downstream field names and risk broken tool chains.
No error handling or recovery guidance. Tools return no error messages, categorization (retryable vs fatal), or suggestions for next steps. An LLM hitting a 404 or timeout has no guidance.
list_airports lacks pagination and result count. No limit parameter, no total count in response. If inventory grows, this tool could return thousands of airports, exhausting context.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2026-07-28+ | v2 |
Search hotels in a city for a date range. When a venue_id is given, each result carries its straight-line distance to that venue, which is the field to rank on when the traveller has meetings there.
No tool annotations despite all tools being read-only. Adding readOnlyHint=true to all 7 tools would signal to agents that these are safe to call speculatively without side effects.
Parameter descriptions lack formal constraints. 'YYYY-MM-DD' is mentioned in text but not declared as a JSON Schema format or pattern. LLMs may pass invalid dates like '2024-13-45'.