MCP resources¶

Beyond the eight tools, Lawbster exposes MCP resources. Resources are URI-addressable content (legal://…) that clients can use as context sources — like file attachments, but server-curated.

Two categories:

Static resources¶

legal://rechtsrahmen¶

Domain → relevant laws. A curated mapping that helps LLMs identify the right legal framework for a question.

Examples:

• Data protection: GDPR, BDSG, TTDSG, AI Act (processing aspects)
• Insurance law: VVG (special statute), BGB (subsidiary) — not just BGB
• Tax law: AO (procedural), EStG / KStG / UStG (substantive) — not BGB

The resource also flags known confusion clusters ("VVG vs BGB", "AO vs EStG"), so the LLM doesn't fall back to general law when a specialised statute applies.

legal://filter_values¶

Valid values for filters in legal_search, legal_list_laws, legal_find_citing_decisions. A bot that knows all valid values upfront avoids INVALID_FILTER errors.

legal://eu_celex_registry¶

Curated CELEX identifiers for major EU acts (GDPR, AI Act, MiCA, NIS2, …). Mapping name → CELEX ID → EUR-Lex URI. Valuable because EUR-Lex's own search is finicky and CELEX IDs are often cited directly.

Dynamic resource templates¶

legal://norm/{norm_id}¶

Full text of a single norm by stable norm ID. Same output as legal_lookup, but resource-addressable — some MCP clients (Copilot Studio, Claude.ai) integrate resources better than tool outputs.

legal://law/{source_type}/{abbreviation}¶

Table of contents of a law as a resource. Same output as legal_get_toc. Example: legal://law/gii/BGB.

legal://eu/celex/{celex}¶

EU act by CELEX ID with TOC and direct article links. Example: legal://eu/celex/32016R0679 (GDPR).

When resources vs. tools?¶

Rule of thumb: static resources are always free and a good default once the client supports them. Dynamic resources cost the same as tools — use whichever your client integrates best.