API playground — verify behavior with anonymous 3 req/day
Try it free with anonymous 3 req/day per IP (no registration, no API key, resets JST next-day 00:00). Call https://api.jpcite.com/v1/* directly from the browser to inspect the Evidence Packet structure, source_url / source_fetched_at / known_gaps, and the professional-practice-law fence envelope. After 3 calls, continue by issuing an API key (fully metered at ¥3.30/unit).
Quick use
- Anonymous calls are limited to 3 requests per day per IP; use an API key for larger evaluations.
- Keep
source_urland_disclaimerfields in sample responses; they are part of the integration contract.
This is the production API. Requests reach api.jpcite.com immediately and consume the anonymous quota (3/day); it resets at JST next-day 00:00. You can check your remaining quota with the indicator on the live page.
3-step Evidence wizard
The live wizard connects to the jpcite playground stream over SSE and renders each step progressively in section chunks; step 1 → step 2 → step 3 advance by button click.
Step 1 / recommended-program search
Enter a corporate number (13 digits) and a confirmation purpose. As an evidence check before handing long materials to an AI, it runs a search equivalent to GET /v1/intelligence/precomputed/query over SSE and returns recommended programs, citations, and Evidence-Packet metadata. By default source_tokens_basis=pdf_pages · source_pdf_pages=30 · input_token_price_jpy_per_1m=300 are filled in, so submitting returns source_url, source_fetched_at, known_gaps, and compression.input_context_reduction_rate. On the second try, change source_pdf_pages to see the difference; on the third, substitute your own question.
Step 2 / combination & exclusivity compatibility check
The recommended programs from step 1 are fed into an equivalent of /v1/compatibility/check, receiving exclusivity / combination rules and copy_paste_parts (200-character application-text templates) over SSE.
Step 3 / evidence-pack generation
Final packaging, equivalent to /v1/evidence/packets/query. Over SSE, work_queue / source_receipts / known_gaps arrive incrementally, and finally next returns the artifact URL. Output types include DD question sheet (30–60 questions), monthly watch registration, and deep program detail.
Tools you can call
| Tool | What it searches |
|---|---|
search_programs | Program search (subsidies / loans / tax / certifications, 11,601) |
search_cases | Case studies (2,286) |
search_loans | Loans (108; collateral / personal-guarantee 3-axis) |
search_laws | Laws (9,484 catalog / 6,493 full text) |
search_tax_rules | Tax rulesets (50) |
check_enforcement | Enforcement actions (1,185) |
A 500 ms debounce means rapid clicks consume only one request. Each call returns the response highlights and response headers for inspection.
Connect from an AI agent
The live page detects connections from AI agents (e.g., Claude MCP) versus a normal browser. Add one line in Claude Desktop / Claude Code, etc., to use 261 MCP tools; you can continue beyond the anonymous quota with an API key.
claude mcp add jpcite -- uvx autonomath-mcpSwitch to continued use
An evidence pack whose structure you verified in the 3 steps can be run continuously at the scale of 100 or 1,000 companies. Pricing is fully metered at ¥3/billable unit tax exclusive (¥3.30 tax inclusive) — no tiers, no fixed monthly fee, stoppable anytime. Issue an API key on the pricing page.
This service is for information retrieval. Consult a qualified professional for individual, case-specific tax or legal determinations.
What comes back
jpcite returns a compact Evidence Packet or output artifact, not a final legal, tax, audit, credit, or application decision. Typical fields include source_url, source_fetched_at, corpus_snapshot_id, known_gaps, matched public records, professional-scope notes, and optional input-context estimates.
For the live interactive playground (SSE wizard), see the Japanese interactive page. For English FDI examples, see jpcite for foreign-invested entities.