For developers — API & MCP

Read 1,295 source-checked records of AI history from your own code and AI tools. It's read-only — no sign-up, no key.

API base URLhttps://next-token.net/api/v1
MCP endpointhttps://next-token.net/mcp
LicenseCC BY-NC 4.0
Rate limitabout 120 a minute

What's available

Read-only API

Search and read records, on this day, AI words, axes and the sentence checker. GET only, JSON answers.

MCP server

Add one address to Claude, ChatGPT, Cursor or VS Code and the assistant can look up records and cite them.

Download the dataset

Need everything? Don't page through the API — download the CSV or JSON files from the data page in one go.

Dataset page

More ways to use it

Today in AI history widget

One line of HTML on a blog or school site shows an AI record from that calendar day, every day.

How to embed it

llms.txt

A plain-text map of this site for AI tools: the key pages, the data, the API and the MCP server.

Open llms.txt

API

The base URL is https://next-token.net/api/v1. Every answer is JSON, and CORS is open (Access-Control-Allow-Origin: *) so any web page can call it.

Common rules

  • GET only — anything else gets 405.
  • Pick the language with lang=ko (default) or lang=en.
  • Read answers carry Cache-Control: public, max-age=300 — don't ask for the same thing again within five minutes.
  • Bad values get 400, missing things 404, too many calls 429 with Retry-After. Errors are JSON too: {"error": {"code", "message"}}.
  • Every answer includes license, attribution and docs. When you show the data, show the attribution line with it.

GET/api/v1

API index — endpoints, parameters and license

Example
curl https://next-token.net/api/v1

GET/api/v1/records

Search and filter records. With q they come in order of relevance (BM25); without it, by date.

Parameters
NameMeaning
qWords to search for (Korean or English, up to 200 characters)
axisAxis number 0–21 (main or secondary axis)
yearYear (e.g. 2017)
fromFrom this date (YYYY-MM-DD)
toUp to this date (YYYY-MM-DD)
sourceprimary (primary source) or secondary (press)
orderasc (default) or desc — date order when there's no q
limitHow many at once (1–50, default 20)
offsetHow many to skip (next page)
langko or en
Example
curl -G https://next-token.net/api/v1/records --data-urlencode "q=transformer paper" -d lang=en -d limit=1
Sample answer (trimmed)
{
  "data": [
    {
      "id": "E0078",
      "url": "https://next-token.net/en/e/E0078/",
      "lang": "en",
      "title": "Google releases the Transformer paper Attention Is All You Need",
      "date": "2017-06-12",
      "date_precision": "day",
      "date_text": "Jun 12, 2017",
      "axis": 1,
      "axis_name": "Prompt Engineering",
      "also_axes": [],
      "actor": "Google (Vaswani et al.)",
      "event": "Released the Transformer paper Attention Is All You Need",
      "significance": "Self-attention alone, without recurrence, made parallel training possible, establishing th…",
      "source": {
        "url": "https://arxiv.org/abs/1706.03762",
        "domain": "arxiv.org",
        "type": "primary",
        "kind": "paper"
      },
      "claim_type": "fact",
      "added_week": "2026-W35",
      "score": 10.71
    }
  ],
  "meta": {
    "total": 70,
    "limit": 1,
    "offset": 0,
    "lang": "en",
    "order": "relevance",
    "query": {
      "q": "transformer paper"
    }
  },
  "license": {
    "name": "CC BY-NC 4.0",
    "url": "https://creativecommons.org/licenses/by-nc/4.0/"
  },
  "attribution": "다음 조각 · Next Token (next-token.net) — factnotes, https://blog.naver.com/factnotes",
  "docs": "https://next-token.net/developers/"
}

GET/api/v1/records/{id}

One record (e.g. E0004)

Parameters
NameMeaning
langko or en
Example
curl https://next-token.net/api/v1/records/E0004?lang=en
Sample answer (trimmed)
{
  "data": {
    "id": "E0004",
    "url": "https://next-token.net/en/e/E0004/",
    "lang": "en",
    "title": "McCarthy, Minsky, Rochester and Shannon write the Dartmouth proposal",
    "date": "1955-08-31",
    "date_precision": "day",
    "date_text": "Aug 31, 1955",
    "axis": 0,
    "axis_name": "Prehistory — From Neural Networks to the Transformer",
    "also_axes": [],
    "actor": "John McCarthy, Marvin Minsky, Nathaniel Rochester, Claude Shannon",
    "event": "Wrote the proposal for the Dartmouth Summer Research Project — the first use of the term \"artificial intelligence\"",
    "significance": "The document that gave the field its name. It led to the workshop the following year, in 1956",
    "source": {
      "url": "http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html",
      "domain": "www-formal.stanford.edu",
      "type": "primary",
      "kind": "official"
    },
    "claim_type": "fact",
    "added_week": "2026-W35"
  },
  "license": …, "attribution": …, "docs": …
}

GET/api/v1/today

On this day — records from the same month and day. Without a date it uses today in Korea (KST).

Parameters
NameMeaning
dateMM-DD (e.g. 03-14)
langko or en
Example
curl https://next-token.net/api/v1/today?date=03-14&lang=en
Sample answer (trimmed)
{
  "data": {
    "date": "03-14",
    "page": "https://next-token.net/en/today/03-14/",
    "records": [
      {
        "id": "E0201",
        "url": "https://next-token.net/en/e/E0201/",
        "lang": "en",
        "title": "Anthropic launches Claude",
        "date": "2023-03-14",
        "date_precision": "day",
        "date_text": "Mar 14, 2023"
      }
    ]
  },
  "meta": {
    "lang": "en"
  },
  "license": …, "attribution": …, "docs": …
}

GET/api/v1/terms

AI Words — the list, or a search

Parameters
NameMeaning
qSearch by word or meaning
langko or en
Example
curl -G https://next-token.net/api/v1/terms --data-urlencode "q=Artificial intelligence" -d lang=en
Sample answer (trimmed)
{
  "data": [
    {
      "slug": "ai",
      "term": "Artificial intelligence",
      "aka": [
        "AI"
      ],
      "short": "Technology that lets computers do things that used to need a human mind.",
      "group": "basics",
      "level": 1,
      "url": "https://next-token.net/en/glossary/ai/"
    }
  ],
  "meta": {
    "total": 1,
    "lang": "en",
    "query": {
      "q": "Artificial intelligence"
    }
  },
  "license": …, "attribution": …, "docs": …
}

GET/api/v1/terms/{slug}

One AI word — meaning, explanation, analogy and related records

Parameters
NameMeaning
langko or en
Example
curl https://next-token.net/api/v1/terms/ai?lang=en

GET/api/v1/axes

The 22 axes — names, introductions and counts

Parameters
NameMeaning
langko or en
Example
curl https://next-token.net/api/v1/axes?lang=en
Sample answer (trimmed)
{
  "data": [
    {
      "axis": 0,
      "mark": "⓪",
      "name": "Prehistory — From Neural Networks to the Transformer",
      "lead": "What piled up over 80 years before AI reached the Transformer?",
      "body": "…",
      "count": 66,
      "url": "https://next-token.net/en/axis/0/"
    }
  ],
  "meta": {
    "total": 22,
    "lang": "en"
  },
  "license": …, "attribution": …, "docs": …
}

GET/api/v1/check

Check a sentence — the same rules as “Is it true?”. The verdict is match, mismatch, partial or not_found.

Parameters
NameMeaning
claimThe sentence to check (up to 300 characters). Never stored or logged.
langko or en
Example
curl -G https://next-token.net/api/v1/check --data-urlencode "claim=The Dartmouth workshop took place in 1956" -d lang=en
Sample answer (trimmed)
{
  "data": {
    "claim": "The Dartmouth workshop took place in 1956",
    "lang": "en",
    "verdict": "match",
    "label": "The year matches the record",
    "message": "Checked against the record “McCarthy, Minsky, Rochester and Shannon write the Dartmouth proposal”: the year in your sentence matches. Only the year was compared, so this doesn't mean the whole sentence is right.",
    "matched": [
      "year"
    ],
    "reason": "same",
    "parts": [
      {
        "part": "year",
        "claim": "1956",
        "record": "1956",
        "ok": true,
        "basis": "record_text"
      }
    ],
    "records": [
      {
        "id": "E0004",
        "url": "https://next-token.net/en/e/E0004/",
        "lang": "en",
        "title": "McCarthy, Minsky, Rochester and Shannon write the Dartmouth proposal",
        "date": "1955-08-31",
        "date_precision": "day",
        "date_text": "Aug 31, 1955"
      }
    ],
    "note": "Not being in the record doesn't make a sentence false. This verdict only compares it with the record."
  },
  "license": …, "attribution": …, "docs": …
}

MCP server

The endpoint is https://next-token.net/mcp. It speaks Streamable HTTP and keeps no state (no sessions). It connects straight away — no sign-in, no key. Supported protocol versions: 2026-07-28 · 2025-11-25 · 2025-06-18 · 2025-03-26 · 2024-11-05. A batch may hold up to 4 messages, and each message counts toward the rate limit.

Connect it

Claude (web and desktop app)

Settings → Connectors → “Add custom connector”. Give it a name (e.g. Next Token) and paste the address below. No authentication needed.

What to add
https://next-token.net/mcp

Claude Code

One line in your terminal.

What to add
claude mcp add --transport http next-token https://next-token.net/mcp

Cursor

Add it to ~/.cursor/mcp.json (all projects) or the project's .cursor/mcp.json.

What to add
{
  "mcpServers": {
    "next-token": {
      "url": "https://next-token.net/mcp"
    }
  }
}

VS Code

Add it to the workspace's .vscode/mcp.json, or run “MCP: Add Server” from the Command Palette, choose HTTP and paste the address.

What to add
{
  "servers": {
    "next-token": {
      "type": "http",
      "url": "https://next-token.net/mcp"
    }
  }
}

ChatGPT (developer mode)

On plans with developer mode: Settings → Apps → Advanced settings → turn on developer mode, then create an app with the address below and choose no authentication.

What to add
https://next-token.net/mcp

Tools

Every tool only reads (readOnlyHint). Each answer comes as short text (with record links and the attribution line) and as structured JSON (structuredContent).

Tools
ToolMeaningArguments
search_recordsSearch records (narrow by axis and years)query, axis?, year_from?, year_to?, lang?, limit?
get_recordOne full recordid, lang?
on_this_dayRecords from the same month and daydate? (MM-DD), lang?
timelineA year's or an axis's records in date orderyear? | axis?, lang?, limit?
check_claimCheck a sentence against the recordclaim, lang?
define_termMeaning and explanation from AI Wordsterm, lang?
citeA citation for a record (plain, APA, BibTeX)id, style (plain | apa | bibtex), lang?

Limits and being gentle

  • About 120 requests a minute per client. Beyond that you get 429 with Retry-After.
  • The server runs on a free tier that allows only a few milliseconds of computing per request; searches use a prebuilt index.
  • If you need all records, download the dataset files; keep answers for five minutes and reuse them.
  • Please don't poll — new records are added once a week.

Sentences sent to the checker (/check, check_claim) are used only to make the answer — never stored or logged. Other requests aren't kept either.

License and attribution

Records and texts are under CC BY-NC 4.0: for non-commercial use you may use, adapt and share them with attribution. Please credit like this:

다음 조각 · Next Token (next-token.net) — factnotes, https://blog.naver.com/factnotes

The full files and their columns are on the dataset page. Dataset page