Loading ToolBench
MCP server for Thailand data: DBD company registry lookup, Thai address normalization, holidays, and VAT/WHT reference
Strong tool definitions with comprehensive descriptions and clear use cases. All 5 tools have detailed, LLM-optimized descriptions (150-400 chars) explaining WHAT, WHEN, and WHY. Input schemas are present with type constraints and patterns. However, output schemas are documented in prose rather than formal JSON Schema, and error handling lacks actionable recovery guidance. Tool names follow verb_noun convention (lookup, normalize, list, reference). Parameter descriptions are thorough with examples and constraints. No critical security issues detected.
Parse free-text Thai addresses (Thai or English) into the official administrative hierarchy: province (จังหวัด) → district (อำเภอ/เขต) → subdistrict (ตำบล/แขวง) → postal code, with bilingual names and official codes. Handles common written forms: abbreviations (ต./อ./จ., กทม.), Bangkok's แขวง/เขต terminology, missing components, and postcode-only inputs. Matching is dictionary-based against the complete official geography (77 provinces, 928 districts, 7,436 subdistricts). Args: - address (string): free-text address, 2-500 chars Returns structured JSON: { "province": { "code": number, "nameTh": string, "nameEn": string } | null, "district": { ... } | null, "subdistrict": { ... } | null, "postalCode": number | null, "confidence": "high" | "medium" | "low" | "ambiguous" | "none", "streetHint": string | null, // residual text after removing administrative parts (house no., road, soi) "warnings": string[], // e.g. postcode/subdistrict mismatch "candidates": [...] // present when confidence = "ambiguous"; add district/postcode and retry } Examples: - Use when: filling a shipping form from a customer's pasted address - Use when: validating that district/province/postcode in a record are consistent - Use when: deduplicating customer records with differently-written addresses - Don't use when: you need geocoding to lat/long (not provided) Error handling: confidence "none" means nothing matched — the text may not be a Thai address.
Look up a Thai company (juristic person) in the official DBD (Department of Business Development) registry by its 13-digit registration number. Returns the official registry profile: Thai/English name, entity type, operating status, registration date, registered and paid-up capital (THB), business objective (bilingual, with TSIC-style code), and registered head-office address. Args: - juristic_id (string): exactly 13 digits, e.g. "0107536000315" Returns structured JSON: { "found": boolean, "company": { "juristicId": string, "nameTh": string, "nameEn": string, "type": string, // e.g. "บริษัทจำกัด" (Co., Ltd.), "บริษัทมหาชนจำกัด" (PCL) "status": string, // e.g. "ยังดำเนินกิจการอยู่" = still operating "registerDate": string, // YYYYMMDD "registeredCapitalBaht": number, "paidUpCapitalBaht": number, "objective": { "code": string, "textTh": string, "textEn": string }, "address": { ... } } | null, "fromCache": boolean, "attribution": string } Examples: - Use when: verifying a Thai supplier/counterparty exists and is still operating (KYB / due diligence) - Use when: fetching official registered capital or address for a contract - Don't use when: you only have a company NAME — this tool requires the 13-digit ID (ask the counterparty or check their invoice/website footer) Notes: - "found": false means the ID is not in the registry — it may be mistyped, or the entity may be a sole proprietor (not a juristic person). - Data is the live official registry, cached for 24h.
Output schemas documented in prose descriptions rather than formal JSON Schema. LLMs cannot parse unstructured schema documentation to plan downstream calls or validate responses.
Error handling lacks actionable recovery guidance. No tool documents what to do when a lookup fails (e.g., 'company not found, try verifying the 13-digit ID format' or 'address confidence=none, provide more components').
thai_vat_reference and thai_list_provinces have no input parameters but descriptions do not explicitly state they are discovery/reference tools with no arguments. Ambiguous for LLM tool selection.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
List Thai public holidays for a given year (2025, 2026), with dates, Thai/English names, and a flag for lunar-calendar holidays. Args: - year (number): one of 2025, 2026 Returns structured JSON: { "year": number, "count": number, "holidays": [{ "date": "YYYY-MM-DD", "nameEn": string, "nameTh": string, "lunar": boolean }], "disclaimer": string } Use when scheduling anything in Thailand: payment due dates, meeting planning, delivery estimates, contract deadlines. Lunar holidays (lunar: true) are fixed by annual announcement — verify for critical deadlines. Bank holidays can differ slightly from public holidays (Bank of Thailand publishes its own list).
List all 77 Thai provinces with official codes and bilingual names. Returns structured JSON: { "count": 77, "provinces": [{ "code": number, "nameTh": string, "nameEn": string }] } Use when you need the canonical province list for validation, dropdowns, or mapping Thai/English province spellings.
Reference card for Thai VAT and withholding tax (WHT) basics: standard VAT rate, registration threshold, filing forms and deadlines, and common WHT rates by payment category (services, rent, advertising, transport, dividends, etc.). No arguments. Returns structured JSON with "vat", "withholdingTax" (rates table + certificate requirement), and "disclaimers". Use when: computing invoice amounts involving Thai WHT deduction, checking whether a business must register for VAT, or explaining the 50 ทวิ withholding certificate. Reference only — NOT tax advice; verify with the Revenue Department for filings.
thai_address_normalize returns 'candidates' array when confidence='ambiguous' but no guidance on how to retry or what additional parameters to provide. Leaves agent guessing.
No pagination or result limits documented for thai_list_provinces (77 items) or thai_holidays (variable count). While current datasets are small, no guidance on future scaling or max result sizes.