tw-law-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_law_packsA | List enabled Taiwan/New Taipei law packs for a jurisdiction, case type, and procedure stage. |
| build_law_snapshotC | Build a versioned law snapshot for a run context and return source-bound entries. |
| search_lawC | Search the deterministic P0 law snapshot. Results include rank, license, source URL, and checksum. |
| get_articleB | Get one article from the P0 law snapshot by law_id and article number. |
| verify_citationB | Verify whether a citation exists in the P0 snapshot. This is deterministic citation existence, not legal assurance. |
| check_claim_supportC | Check whether a claim is sufficiently supported by article text for fail-closed filtering. |
| get_local_ruleB | Get lifecycle-aware structured local-law metadata; inactive, pending, ambiguous, malformed, missing-lifecycle, or historically indeterminate rules fail closed. |
| detect_illegal_construction_referenceC | Detect whether files or text only imply illegal-construction indicators for manual routing. |
| get_source_policyC | Return authority rank, license status, update policy, and crawl policy for a source URL. |
| compare_source_policiesB | Return official-source authorization and update-policy differences for P0 sources. |
| run_source_policy_acceptanceB | Verify P0 article sources have complete official-source policy evidence and comparison coverage. |
| list_jurisdictionsC | List enabled jurisdiction registry entries, optionally including disabled stubs. |
| run_jurisdiction_registry_acceptanceB | Verify jurisdiction registry entries are enabled or fail-closed with law-pack coverage. |
| run_packaging_acceptanceA | Verify Codex and Claude Code wrappers preserve standalone MCP server packaging strategy. |
| run_phase_acceptanceB | Run aggregate acceptance for all roadmap Phase gates, including local-rule lifecycle validation. |
| run_scenario_matrix_acceptanceB | Verify Taiwan scenario matrix fixtures declare corpus packs, tool boundaries, artifacts, gates, and HITL policies. |
| run_data_layout_acceptanceB | Verify split source packs, registries, fixtures, and normalized source_unit contracts. |
| run_source_adapter_acceptanceB | Verify deterministic source adapters produce normalized source_units. |
| resolve_tw_scenarioD | Resolve Taiwan/New Taipei interior renovation scenario routing to source packs, artifacts, and gates. |
| check_fire_equipment_routingD | Fail-closed routing for fire-safety-equipment document/professional confirmation needs. |
| check_fire_compartment_evidenceC | Find fire-compartment related evidence terms and source-bound human-confirmation needs. |
| check_material_evidenceC | Check material evidence metadata presence without judging material authenticity. |
| build_ntpc_submission_packetC | Build New Taipei submission/completion packet checklist for a procedure stage. |
| plan_web_search_fallbackC | Return an official-source fallback plan for corpus misses without answering from live search. |
| resolve_procedure_requirementsC | Return stage-specific New Taipei interior renovation procedure requirements. |
| resolve_procedure_stage_confidenceB | Score procedure_stage confidence from document text and file metadata; low confidence routes to HITL. |
| get_fixture_baseline_statusB | Report whether the de-identified fixture baseline satisfies G2. |
| run_fixture_pipeline_acceptanceB | Run the synthetic G2 fixture baseline through snapshot, sheet, HITL, and audit-gate acceptance. |
| extract_file_metadataC | Extract metadata-only file contract fields without allowing raw drawing/document content into agent input. |
| parse_masked_documentC | Parse masked Taiwan official-document text into document_parsed fields and procedure-stage signal. |
| normalize_atomic_correction_itemsC | Normalize parsed document sections into atomic correction items with source spans. |
| build_sheet_manifestC | Build sheet_manifest metadata from drawing/source snapshot filenames without reading raw pixels. |
| build_hitl_confirmation_packetC | Build client_questions for low-confidence procedure stage and manual-review correction items. |
| apply_hitl_confirmationsA | Apply human answers to procedure-stage and correction-item HITL questions with fail-closed validation. approval_provenance.approval_status stays 'unapproved' unless an authenticated approval was recorded server-side for the run. |
| run_audit_gatesC | Run deterministic audit gates (schema/citation/source/claim/redline/governance). |
| run_source_coverage_acceptanceA | Verify the law corpus against the source packs: every pack-referenced article exists, every article has a source policy, and pending articles carry no unverified text. |
| run_tw_corrections_analysisB | Run stage 1 Taiwan corrections analysis from masked text and metadata-only files. |
| run_tw_corrections_responseA | Run stage 2 Taiwan corrections response from analysis artifacts and human answers. Artifacts are verified against the server-issued run digest; caller-asserted confirmation fields are ignored and unapproved answers fail closed. |
| run_two_stage_flow_acceptanceC | Verify deterministic two-stage Taiwan contractor flow artifacts and red-line policy. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 39 tools
There is substantial overlap among the many run_*_acceptance tools (e.g., run_source_policy_acceptance vs run_source_coverage_acceptance) and among check_* routing tools, making it difficult to select the correct one without reading details. While descriptions clarify intent, the boundaries between acceptance, audit, and analysis tools are blurry.
Names consistently follow a verb_noun snake_case pattern with predictable subpatterns like run_*_acceptance, check_*, and build_*. Minor deviations exist (e.g., run_audit_gates vs run_*_acceptance, get_fixture_baseline_status vs other get_* tools) but the overall style is coherent.
With 39 tools, the surface is heavily overloaded. Many tools are internal acceptance/QA checks (run_*_acceptance, run_audit_gates) that are unlikely to be needed by end users, plus the core business tools, resulting in excessive cognitive load for an agent.
The tool set covers an end-to-end workflow: law snapshot/search/article retrieval, citation and claim verification, scenario routing, document parsing, correction normalization, HITL confirmations, audit gates, and acceptance validation. Minor gaps exist in management/write operations (e.g., updating law packs or jurisdiction entries), but these are not core to the apparent read/analysis-focused purpose.