jpcite — Evidence-first context layer for Japanese public-program data

jpcite — Evidence-first context layer for Japanese public-program data

mcp-name: io.github.shigetosidumeda-cyber/autonomath-mcp

Current public docs and release tags are the source of truth for version and pricing.

LLM agent / RAG パイプラインに渡す前の compact context を REST + MCP で返します。公開行には 出典 URL + content_hash + 取得日時 が付き、官公庁・自治体・公庫・公式事業者ページなど確認可能な出典を優先します。11,684 programs / 9,484 laws / 1,185 enforcement cases + 22,258 enforcement-detail records / 96 MCP tools / median 7 day freshness。LLM は呼び出さず、民間まとめサイトにも依存しません。通常の検索・取得は ¥3/billable unit、anon 3/日 free。

English: Evidence-first context layer for Japanese public-program data, exposed as REST + MCP. Published rows return source URL + content_hash + fetched_at so an LLM agent or RAG pipeline can ground answers on verifiable official sources — no LLM calls inside the service, no aggregator scraping. 11,684 programs / 9,484 laws / 1,185 enforcement cases + 22,258 enforcement-detail records / 96 MCP tools / median 7 day freshness. ¥3/billable unit for normal search/detail calls, 3/day free anonymous.

Use jpcite when an AI answer needs Japanese public-program evidence, source URLs, fetched_at metadata, compatibility rules, enforcement checks, or known gaps before drafting prose. Skip it for short general questions, translation, brainstorming, or topics that do not need source-linked Japanese institutional data.

Token and cost impact is workload-dependent. jpcite can reduce the input context a caller sends to GPT/Claude by returning compact source-linked facts first, but it is not a provider billing guarantee; output tokens, reasoning tokens, tool calls, search, cache behavior, and model choice remain controlled by the caller.

How jpcite compares to single-source MCP servers

jpcite is the 横断 + Evidence Packet layer. The 3 active single-source Japanese MCP servers each handle one slice — they are complementary, not competitive:

• vs jgrants-mcp (digital-go-jp/jgrants-mcp-server, 5 tools, jGrants 補助金 only): jpcite adds 法令 / 判例 / 行政処分 / 適格事業者 / 法人 360° / 排他併用判定. Use jgrants-mcp for the grant application path; use jpcite for cross-source compliance check. → /compare/jgrants-mcp/
• vs tax-law-mcp (kentaroajisaka/tax-law-mcp, 7 tools, e-Gov + NTA + KFS live scrape): jpcite adds 50 structured tax_rulesets + 9,484 e-Gov laws + 28,201 article rows pre-indexed (median <100ms, no live-scrape latency) + 通達 cross-ref to 制度 / 採択 / 行政処分. Use jpcite for pre-indexed answers + 通達 cross-ref; use tax-law-mcp for ad-hoc lookups. → /compare/tax-law-mcp/
• vs japan-corporate-mcp (yamariki-hub/japan-corporate-mcp, 8 tools, gBizINFO + EDINET + e-Stat live API, 3 user keys required): jpcite ships pre-indexed 166,969 法人 + 13,801 適格事業者 + 1,185 行政処分 + 22,258 enforcement detail with anonymous trial (no user API key required). Use jpcite for analyst pre-screening; use japan-corporate-mcp for live regulator pulls when keys are already provisioned. → /compare/japan-corporate-mcp/

What this is

An evidence-first context layer over Japanese institutional public data, exposed as REST + MCP. Published rows carry a source URL, a content_hash, and a fetched_at timestamp so downstream LLM agents or RAG pipelines can cite back to verifiable official source pages without re-crawling.

What this isn't

• Not legal advice (弁護士法 § 72)
• Not tax advice (税理士法 § 52)
• Not 行政書士 work (行政書士法 § 1)
• Not real-time amendment tracking (snapshot data, partial historical diffs)
• Verify primary sources before any business decision

Coverage

• Source-linked records — most published rows include source_url, content_hash, and source_fetched_at; known source gaps are surfaced. Known second-tier aggregator pages are excluded from citation sources where detected.
• 11,684 searchable programs across 47 prefectures + national (補助金・融資・税制・認定; tier S=114 / A=1,340 / B=4,186 / C=6,044; full catalog = 14,472, 2,788 publication-review rows)
• 2,286 採択事例 + 108 融資 (担保・個人保証人・第三者保証人 三軸分解) + 1,185 行政処分 + 22,258 enforcement-detail records + 2,065 court decisions + 362 bids
• 154 laws full-text indexed + 9,484 law metadata records (e-Gov CC-BY; full-text coverage is incremental — name resolver covers all 9,484, body text 154) + 50 tax rulesets + 13,801 invoice registrants (PDL v1.0 delta)
• 181 exclusion / prerequisite rules (125 exclude + 17 prerequisite + 15 absolute + 24 other) — surfaced as structured eligibility predicates, not free-text
• 96 MCP tools in the standard public configuration (39 core + 3 audit/composition + 25 jpcite generic + 4 universal + 5 static-resource tools + 4 NTA corpus + 10 composition tools + 3 industry-pack tools + 3 corporate-layer tools), protocol 2025-06-18, stdio. Optional labor-agreement tools are disabled unless explicitly enabled.
• REST API — endpoints under /v1/programs/*, /v1/laws/*, /v1/tax_rulesets/*, /v1/case-studies/*, /v1/loan-programs/*, /v1/enforcement-cases/*, /v1/exclusions/*, /v1/am/*. OpenAPI: docs/openapi/v1.json
• No LLM inside the service — no external LLM calls in the data/evidence path. Content endpoints are generated from the corpus and deterministic application code; reasoning lives in the caller's agent.
• Median 7-day freshness on tier S/A program rows; per-source source_fetched_at distribution exposed at GET /v1/stats/freshness
• ¥3/billable unit metered (tax-exclusive; 税込 ¥3.30). Normal search/detail calls are 1 unit; batch/export endpoints document their formula. Anonymous 3 req/日 free (no signup; JST 翌日リセット)

30-second quickstart (Claude Desktop)

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "jpcite": {
      "command": "uvx",
      "args": ["autonomath-mcp"]
    }
  }
}

Restart Claude Desktop, then ask: 「東京都で設備投資に使える補助金を教えて。」

HTTP fallback (uvx インストール時)

uvx autonomath-mcp で取得した wheel には DB が同梱されていないため、起動時にローカル DB が空であることを検知し、自動で api.jpcite.com への HTTP fallbackモードに切替えます。匿名 3 req/日 は IP 単位で同一に適用し、paid key は ¥3/billable unit でメータリングします。

{
  "mcpServers": {
    "jpcite": {
      "command": "uvx",
      "args": ["autonomath-mcp"],
      "env": {
        // optional: jpcite API key (匿名 3 req/日 を超える場合)
        // This is sent to api.jpcite.com as X-API-Key; it is not an LLM key.
        "JPCITE_API_KEY": "am_xxx",
        // optional: custom upstream
        "JPCITE_API_BASE": "https://api.jpcite.com"
      }
    }
  }
}

このキーは jpcite API 用です。jpcite サービス内で OpenAI / Anthropic / Gemini などのLLM API は呼びません。

HTTP fallback で完全に動作するツール (top 10): search_programs / get_program /search_case_studies / search_loan_programs / search_enforcement_cases /search_tax_incentives / search_certifications / list_open_programs /dd_profile_am (REST chain hint) / rule_engine_check (remote_only)。上記以外の default-on MCP tools は error: "remote_only_via_REST_API" を返し、対応する REST URL を案内します。フル機能を使う場合はリポジトリを clone してローカル DB を取得してください。

30-second quickstart (REST)

# Primary (X-API-Key header, used across our docs)
curl "https://api.jpcite.com/v1/programs/search?q=設備投資&prefecture=東京都" \
  -H "X-API-Key: am_xxx"

# Also supported: Bearer token
curl "https://api.jpcite.com/v1/programs/search?q=設備投資&prefecture=東京都" \
  -H "Authorization: Bearer am_xxx"

Get an API key at https://jpcite.com/dashboard.

Output sample

GET /v1/programs/search?q=設備投資&prefecture=東京都 returns (truncated to 1 result):

{
  "total": 47,
  "results": [
    {
      "unified_id": "UNI-example-energy-dx",
      "primary_name": "東京都 中小企業 省エネ設備導入支援",
      "amount_max_man_yen": 500,
      "application_window": {"end_date": "2026-06-30"},
      "source_url": "https://www.metro.tokyo.lg.jp/.../energy-dx.html",
      "source_fetched_at": "2026-04-30T00:00:00+09:00",
      "tier": "A"
    }
  ]
}

MCP tools

96 tools at default gates, MCP protocol 2025-06-18, FastMCP over stdio. 完全なリストと引数は docs/mcp-tools.md を参照 (Single source of truth)。

Full list: docs/mcp-tools.md

REST API & SDKs

WARNING: The MCP package is published on PyPI; REST SDKs remain pre-release.

OpenAPI spec

• Agent-safe import: https://api.jpcite.com/v1/openapi.agent.json (docs/openapi/agent.json) for ChatGPT Custom GPT Actions and AI tool importers.
• Full developer spec: https://api.jpcite.com/v1/openapi.json (docs/openapi/v1.json) for SDK generators, Postman, and complete REST reference.

Python MCP package (autonomath-mcp) — package name is kept for client compatibility:

pip install autonomath-mcp

TypeScript / JavaScript SDK (@autonomath/sdk) — package name is kept for compatibility. Public package release is pending; the REST API v1 surface is the stable contract while the SDK remains pre-release.

The package ships dual ESM + CJS output with .d.ts and exposes both REST (@autonomath/sdk) and MCP (@autonomath/sdk/mcp) entry points. Zero runtime dependencies (uses platform fetch).

Runnable examples

• Python: examples/python/ — search by prefecture, check exclusions, program detail, pandas CSV export
• TypeScript: examples/typescript/ — search, exclusions, MCP CLI, Next.js page

Install (Python)

pip install autonomath-mcp
# or
uvx autonomath-mcp

Data sources

Program records prioritize primary sources such as 経産省, 農林水産省 (MAFF), 日本政策金融公庫 (JFC), 総務省, and 47 都道府県公報. Most public records carry source_url + source_fetched_at lineage, with known gaps documented. Public structured data is exposed through /structured/, sitemap-structured.xml, and inline JSON-LD on generated program pages.

Evaluation

Tool quality is publicly verifiable: see evals/ for a 79-query gold-standard suite (gold.yaml + run.py) covering 農業 / 製造 / IT / 創業 / 都道府県 / 税制 / 融資 / 採択事例 / prescreen / 行政処分 / cross-dataset / edge cases / 7 one-shot discovery tools (smb_starter_pack / deadline_calendar / subsidy_combo_finder / similar_cases / subsidy_roadmap_3yr / regulatory_prep_pack). Every expected_ids list was generated against the local evaluation snapshot; CI runs the suite on every PR. Per-tool precision table: see docs/per_tool_precision.md. Run locally with .venv/bin/python evals/run.py.

Self-serve dashboards & transparency

• Dashboard (authenticated): GET /v1/me/dashboard — month-to-date spend, request count, cap state, top tools. See docs/dashboard_guide.md.
• Amendment alerts: POST /v1/me/alerts/subscribe — subscribe by tool / law_id / program_id / industry_jsic / all, with severity gating (critical / important / info). See docs/alerts_guide.md.
• Stats (public transparency): GET /v1/stats/coverage (per-prefecture / authority / kind program counts), GET /v1/stats/freshness (per-source source_fetched_at distribution), GET /v1/stats/usage (anonymised request volume).

Pricing — metered units, no tiers

• ¥3 per billable unit (税込 ¥3.30) — normal search/detail calls are 1 unit, while batch/export endpoints bill by documented fan-out units
• First 3 requests/day free (anonymous, IP-based, JST daily reset)
• No subscription tiers, no seat fees, no annual minimums; anonymous trial calls do not require signup and remain capped at 3 requests/day per IP.
• Cost preview is an estimate, not a billing guarantee — use /v1/cost/preview for jpcite billable-unit estimates. Use evidence packet include_compression=true for caller-supplied input-context comparisons. Provider output/reasoning/search/cache costs remain outside jpcite.

Optional disabled domains

The standard distribution exposes 96 tools for Japanese public-programsearch, evidence, provenance, tax rulesets, laws, court decisions, bids,invoice registrants, and related entity facts. Additional domain-specificsurfaces are intentionally disabled unless enabled through support-managedfeature flags.

• Labor-agreement renderers are disabled by default and are not part of thepublic tool surface.
• Healthcare and real-estate datasets are disabled by default until theirprimary-source coverage and disclaimers are ready for public use.
• Experimental reasoning tools are disabled by default; production callsshould use the documented search, evidence, provenance, and rule-check tools.

Use docs/mcp-tools.md for the current public toolcatalogue and docs/honest_capabilities.mdfor capability boundaries.

SLA & infrastructure

• Monthly uptime target: 99.0% on api.jpcite.com (Fly.ioTokyo + Cloudflare Pages + Cloudflare WAF). Token-bucket rate-limitmiddleware + WAF managed-ruleset are in front of every request.See docs/sla.md.
• Tokushoho disclosure — full statutory disclosure under 特定商取引法at site/tokushoho.html.
• Spec surfaces — site/llms.txt and site/llms-full.txt (JA);site/llms.en.txt and site/llms-full.en.txt (EN) for AI-agentdiscovery.

Support

• Docs: https://jpcite.com/docs/ (search: built-in lunr; Algolia DocSearch integration pending OSS-program approval)
• Email: [email protected]

License

MIT © 2026 jpcite

Keywords: mcp, mcp-server, mcp-tools, claude, rag, agent-tools, japan, japanese, legal-tech, subsidies, grants, loans, tax, tax-incentives, corporate-registry, enforcement, evidence, citation, government, compliance, jpcite, autonomath-mcp, 補助金, 助成金, 融資, 税制優遇, 認定制度, 採択事例, 行政処分, 国税庁, e-Gov, mcp-2025-06-18

Badges

Offline / mirrored copies of the same badges live in badges/for use in environments where shields.io is unreachable.