searchReturn cited corpus results by id and per-part evidence assessment; use fetch for full source text. Query embeddings use Alibaba; candidate passages go to Jina and query plus selected passages to Anthropic.
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes |
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Changes observed during successful MCP inspections.
Input schema / properties / query / descriptionRemoved value: -"English, in the legal register of the act. A non-English query is accepted but retrieves poorly, and the result then carries a `notice` saying so.\nAct identifiers stay verbatim (\"ESPR\", \"2024/1781\"): they are\nmatched as metadata. Article and annex numbers are matched in\ntheir English form: \"第9条\" and \"Artikel 9\" do not match,\n\"Article 9\" does.\nMeasured 2026-08-27 on the live index, one question three ways:\n\"must a marketplace show the product passport before purchase,\nwhere is it written?\" returned NOTHING; \"marketplace must make the\ndigital product passport accessible before purchase\" returned the\nCWA guidance (0.96) with the ESPR at 0.001; \"make the digital\nproduct passport accessible to customers before they are bound by a\ncontract of sale\" returned the ESPR itself first (0.94, p.21)."Input schema / properties / query / descriptionPrevious value: -"ENGLISH ONLY, in the legal register of the act. A non-English `query` violates this tool's contract and its passages must be discarded, not answered from.\nAct identifiers stay verbatim (\"ESPR\", \"2024/1781\"): they are\nmatched as metadata. Article and annex numbers must be rewritten\ninto their English form, because that is how they are matched:\n\"第9条\" and \"Artikel 9\" do not match, \"Article 9\" does.\nMeasured 2026-08-27 on the live index, one question three ways:\n\"must a marketplace show the product passport before purchase,\nwhere is it written?\" returned NOTHING; \"marketplace must make the\ndigital product passport accessible before purchase\" returned the\nCWA guidance (0.96) with the ESPR at 0.001; \"make the digital\nproduct passport accessible to customers before they are bound by a\ncontract of sale\" returned the ESPR itself first (0.94, p.21)."New value: +"English, in the legal register of the act. A non-English query is accepted but retrieves poorly, and the result then carries a `notice` saying so.\nAct identifiers stay verbatim (\"ESPR\", \"2024/1781\"): they are\nmatched as metadata. Article and annex numbers are matched in\ntheir English form: \"第9条\" and \"Artikel 9\" do not match,\n\"Article 9\" does.\nMeasured 2026-08-27 on the live index, one question three ways:\n\"must a marketplace show the product passport before purchase,\nwhere is it written?\" returned NOTHING; \"marketplace must make the\ndigital product passport accessible before purchase\" returned the\nCWA guidance (0.96) with the ESPR at 0.001; \"make the digital\nproduct passport accessible to customers before they are bound by a\ncontract of sale\" returned the ESPR itself first (0.94, p.21)."Input schema / properties / query / descriptionPrevious value: -"The regulatory question — SEND IT IN ENGLISH even if the user\nasked in another language, whatever that language is (the corpus is\nEnglish; a non-English query can miss the official text entirely),\nthen answer the user in their own language. Do NOT translate\nword-for-word: phrase it in the LEGAL REGISTER of the act, because\nretrieval matches the wording of the law, not the phrasing of the\nquestion. Measured 2026-08-27 on the live index, same question three\nways: \"must a marketplace show the product passport before purchase,\nwhere is it written?\" returned NOTHING; \"marketplace must make the\ndigital product passport accessible before purchase\" returned the\nCWA guidance (0.96) with the ESPR at 0.001; \"make the digital\nproduct passport accessible to customers before they are bound by a\ncontract of sale\" returned the ESPR itself first (0.94, p.21). Keep\narticle/annex numbers and act identifiers verbatim (\"Article 9\",\n\"ESPR\", \"2024/1781\"): those are matched as metadata, not translated."New value: +"ENGLISH ONLY, in the legal register of the act. A non-English `query` violates this tool's contract and its passages must be discarded, not answered from.\nAct identifiers stay verbatim (\"ESPR\", \"2024/1781\"): they are\nmatched as metadata. Article and annex numbers must be rewritten\ninto their English form, because that is how they are matched:\n\"第9条\" and \"Artikel 9\" do not match, \"Article 9\" does.\nMeasured 2026-08-27 on the live index, one question three ways:\n\"must a marketplace show the product passport before purchase,\nwhere is it written?\" returned NOTHING; \"marketplace must make the\ndigital product passport accessible before purchase\" returned the\nCWA guidance (0.96) with the ESPR at 0.001; \"make the digital\nproduct passport accessible to customers before they are bound by a\ncontract of sale\" returned the ESPR itself first (0.94, p.21)."Input schema / properties / query / descriptionPrevious value: -"The regulatory question — SEND IT IN ENGLISH even if the user\nasked in another language (the corpus is English; a non-English\nquery can miss the official text entirely), then answer the user in\ntheir own language. Keep article/annex numbers and act identifiers\nverbatim (\"Article 9\", \"ESPR\", \"2024/1781\"): those are matched as\nmetadata, not translated."New value: +"The regulatory question — SEND IT IN ENGLISH even if the user\nasked in another language, whatever that language is (the corpus is\nEnglish; a non-English query can miss the official text entirely),\nthen answer the user in their own language. Do NOT translate\nword-for-word: phrase it in the LEGAL REGISTER of the act, because\nretrieval matches the wording of the law, not the phrasing of the\nquestion. Measured 2026-08-27 on the live index, same question three\nways: \"must a marketplace show the product passport before purchase,\nwhere is it written?\" returned NOTHING; \"marketplace must make the\ndigital product passport accessible before purchase\" returned the\nCWA guidance (0.96) with the ESPR at 0.001; \"make the digital\nproduct passport accessible to customers before they are bound by a\ncontract of sale\" returned the ESPR itself first (0.94, p.21). Keep\narticle/annex numbers and act identifiers verbatim (\"Article 9\",\n\"ESPR\", \"2024/1781\"): those are matched as metadata, not translated."Input schema / properties / query / descriptionPrevious value: -"The regulatory question, in any language (English works best)."New value: +"The regulatory question — SEND IT IN ENGLISH even if the user\nasked in another language (the corpus is English; a non-English\nquery can miss the official text entirely), then answer the user in\ntheir own language. Keep article/annex numbers and act identifiers\nverbatim (\"Article 9\", \"ESPR\", \"2024/1781\"): those are matched as\nmetadata, not translated."Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, idempotentHint, and destructiveHint false, so the safety profile is covered. The description adds valuable behavioral context by revealing the processing pipeline: Alibaba embeddings, Jina candidate passages, and Anthropic for the query and selected passages. This goes beyond the annotations without contradicting them.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences with no filler: the first front-loads the core return value and directs to fetch for full text; the second compactly summarizes the model pipeline. Every clause earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description covers the core purpose and mentions the output is per-part evidence assessment, and an output schema exists to detail return values. However, it omits any comparison with the highly related sibling search_eu_textile_regulations, which is a significant contextual gap for correct tool selection in a sibling-rich environment.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With schema description coverage at 0%, the description must compensate for the undocumented 'query' parameter. It partially does by explaining that the query is embedded with Alibaba and later sent to Anthropic, but it doesn't specify query format, length, supported languages, or example forms. The semantic addition is real but thin.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool returns cited corpus results by id with per-part evidence assessment, and explicitly contrasts itself with fetch for full source text. It doesn't disambiguate from the sibling search_eu_textile_regulations, so it misses full sibling differentiation.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It provides an explicit alternative ('use fetch for full source text') and clearly implies search should be used when cited results or evidence assessments are needed. However, it gives no guidance on when to choose this tool over search_eu_textile_regulations, leaving a gap in the sibling context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.