MCP server for Turkish legislation via the Adalet Bakanligi Bedesten API — fetch legislation with resolvable identifiers and verifiable citations.
Five tools with complete schemas, clear descriptions, and proper error handling. All tools are read-only with proper annotations. Naming follows verb_noun convention (tr_search_*, tr_get_*, tr_list_*, tr_coverage). Descriptions are substantive (100-200 chars typical). Parameters have types and descriptions. Main gaps: no pagination support in search results (capped at 20 hardcoded), parameter descriptions could be more prescriptive about format/constraints, and output schemas are not formally documented in the code.
Declare what this connector covers, how it is sourced, and what it does not cover.
Fetch the full text of a Turkish legislative document by its mevzuat_id.
Fetch the article tree (madde agaci / table of contents) for a Turkish legislative document by its mevzuat_id.
Enumerate all Turkish legislation types with live document counts.
Search Turkish legislation by title and/or full text.
Search results hardcoded to 20 items with no pagination support. Large result sets will be truncated silently, forcing agents to retry with narrower queries. No limit/offset or cursor parameters.
Output schemas not formally documented in code. While responses are structured (SearchResult, LegislationContent, ArticleTree models exist), the tool definitions do not expose return type schemas to the MCP client. LLMs cannot plan downstream calls without seeing what fields are returned.
Parameter descriptions lack prescriptive format guidance. 'mevzuat_turu' says 'Optional legislation type filter, e.g. "KANUN"' but does not state whether values are case-sensitive, what the full enum is, or whether partial matches work. Agents must guess or call tr_list_legislation_types first.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2026-07-28+ | v2 |
tr_coverage tool has minimal description ('Declare what this connector covers...') and no parameters. Its purpose is clear from context but the description is under 100 chars and does not explain when/why an agent should call it relative to other tools.
Error handling returns structured ToolError with [code] prefix, but error messages are generic ('No document found in the Bedesten repository'). When a search returns zero results, the agent is not guided to call tr_coverage or try alternative queries.