Use case · AI
Giving AI agents Japanese equity data
Key points
- check_circleThe MCP server uses the same authentication and quota as the REST API, and returns the same response shapes.
- check_circleYou can check sources and as-of times, so AI answers can cite their basis.
- check_circleAn AI you yourself use is Basic Use. Offering the data to your service's users through AI requires Business or higher.
Before, and with ZENITH Code
Before
- removeAnswers built on web search results had unclear dates and sources for their numbers.
- removeAI reading earnings PDFs misread figures in tables.
- removeUnits (yen, millions, %) were inconsistent and needed conversion before any calculation.
With ZENITH Code
- check_circleThe AI picks from 72 tools and calls the data it needs directly.
- check_circleResponses use consistent units: amounts in yen, ratios as 0–1 decimals.
- check_circle/stable/system/freshness lets you check data freshness before answering.
Endpoints to use
Checks to run before answering, plus common lookups. Each MCP tool calls the matching endpoint.
- GET
/stable/system/freshnessData freshnessCheck data freshness before answering (no key, no tokens)Does not use tokens - GET
/stable/search-nameSearch companies by nameIdentify the securities code from a company name1 tokens per call - GET
/stable/income-statementIncome statementEarnings figures with sources1 tokens per call - GET
/stable/industry-benchmarksIndustry benchmarks by sectorA reference for judging high or low within a sector10 tokens per call - GET
/stable/events-feedCorporate events feedRecent events in one call10 tokens per call
Plan fit and what you may show
Use from an AI agent you yourself use is Basic Use and works on every plan.
| Plan | Monthly | Monthly tokens | Showing to third parties (excluding exchange data) |
|---|---|---|---|
| FreeFits this use | $0 | 5,000 | Not allowed (your own research and analysis, or internal use for a company) |
| ProFits this use | $89 | 100,000 | Not allowed (your own research and analysis, or internal use for a company) |
| BusinessFits this use | $499 | 1,000,000 | Allowed (display in your service, providing processed results, sharing within your organization) |
| Enterprise | Custom | Custom | Allowed (display in your service, providing processed results, sharing within your organization) |
Letting an AI refer to the data at the request of your service's users requires Business or higher. Exchange data such as stock prices may not be given to third parties, including in AI responses.
The Terms are the authoritative source on what you may show or provide. If your use is unclear, contact us before you start. Terms, Article 8 · Pricing
Code example
Call one MCP tool directly (cURL), and fetch the same data through REST (Python).
# Call one MCP tool directly (JSON-RPC over Streamable HTTP)
curl -s https://mcp.zenithcode.net/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_industry_benchmarks","arguments":{"sector":"3650"}}}'
# Check data freshness before answering (no key needed, no tokens used)
curl -s https://api.zenithcode.net/stable/system/freshnessThings to note
- errorMCP authenticates with your API key (Authorization: Bearer or x-api-key header) or with OAuth (browser sign-in) from Claude or ChatGPT connectors.
- errorA tool call uses the same tokens as the endpoint it calls.
- errorRatios are 0–1 decimals. Tell the AI to multiply by 100 when answering in percent.
FAQ for investors and developers
Which AI clients work?expand_more
MCP-compatible clients such as Claude Code, Claude Desktop, Cursor and VS Code. Setup examples are on the MCP guide page.
Can I count the tokens an AI used?expand_more
/stable/usage shows this month's token use by endpoint and by day.
Does it work on the Free plan?expand_more
Yes. MCP tools draw on the same quota as the REST API.
FAQ for AI builders and researchers
How should the AI cite sources?expand_more
Tell it to include the endpoint name and the date fetched. For income figures, the provenance source (TDnet or EDINET) and document ID are also available.
How should the AI treat null?expand_more
null means "not disclosed" or "cannot be computed", not zero. Tell it never to calculate with null as 0.
What if a response is too large?expand_more
Reduce limit, or use filtering parameters such as category and tag in financial-facts.
Disclaimer
This page explains how to use a Japanese equity data API and how its data is defined. It is not a solicitation to invest or a recommendation to buy or sell any security. The content reflects information at the time of writing and is not guaranteed to be accurate or complete. Investment decisions are your own responsibility.
What you may show or provide to third parties depends on your plan and Article 8 of the Terms. Exchange data such as stock prices may not be shown to third parties on any plan. Terms, Article 8 · Data sources and terms of use