Loading ToolBench
MCP server for US law: bills (Congress.gov), US Code (GovInfo), Federal Register, case law (CourtListener), eCFR - with verifiable citations.
Strong domain-specific tool set with clear naming (verb_noun pattern: us_search_*, us_get_*, us_list_*) and comprehensive descriptions (avg 180 chars). All 11 tools have documented input schemas with typed parameters and descriptions. Output schemas are partially documented via citations module. Key gaps: no explicit error handling guidance in descriptions, no parameter constraints (enums for bill_type, doc_type, etc. are mentioned in text but not formalized in schema), and no pagination metadata (limit/offset) in responses. Tool composition is excellent, each tool has single responsibility and clear chaining (search → get pattern). Security is strong: no credentials in parameters, read-only operations only.
Declare what this connector covers, how it is sourced, and what it does not cover - including known gaps and fallback sources.
Full detail for one bill by congress, bill_type, number, including its latest_action_text (current status).
One opinion cluster by cluster_id (take it from us_search_case_law).
Amendment history of one CFR section by title, part, section (e.g. 15, '744', '744.3').
Full metadata + content-format download links for one package by its package_id.
Full metadata for one FR document by its document_number (e.g. '2026-13726'), including the official FR citation and a raw-text URL.
Parameter constraints not formalized as enums in schema. bill_type, doc_type, presidential_document_type, and court accept known finite sets but are defined as free-form strings. LLMs may hallucinate invalid values.
No explicit error handling guidance in tool descriptions. Descriptions do not explain what happens on API failures, rate limits, or empty results. LLMs lack recovery paths.
Pagination metadata incomplete. Tools accept 'limit' but do not document total_count, next_cursor, or offset in responses. Large result sets may exceed context window without clear continuation guidance.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | <=2025-11-25 | v2 |
List enacted-law packages in a GovInfo collection (e.g. 'USCODE', 'CFR', 'FR' for Federal Register) modified since a given date.
List bills of a given congress and bill_type (e.g. 'hr' = House bill, 's' = Senate bill, 'hres', 'sres', 'hjres', 'sjres', 'hconres', 'sconres').
Full-text search over 8M+ US court opinions; use court to scope (e.g. 'cal' = California Supreme Court, 'ny' = NY Court of Appeals, 'scotus'). STATE case law is the reason this tool exists here.
Full-text search over the CURRENT CFR (eCFR, amended through days ago - unlike the GovInfo annual editions).
Full-text search over Federal Register documents; filter by doc_type ('RULE', 'PRORULE', 'NOTICE', 'PRESDOCU') and/or presidential_document_type ('executive_order', 'proclamation', ...) for executive orders and other presidential documents.
Output schemas not explicitly documented in tool definitions. Descriptions reference 'latest_action_text', 'human_readable_citation', 'source_url' but these are not formally declared in input/output schema blocks.
Rate limit guidance incomplete. us_search_case_law mentions '~5 requests/min' in instructions but tool description does not warn about rate limits or provide backoff guidance.