---
name: elements-live
description: Research-intensity data for the 118 chemical elements and thousands of research topics, from Elements Live (elements-live.com). Use when a question is about how much scientific research activity an element, material or topic attracts, whether that activity is rising or fading, how two elements compare, or which papers to read on an element or topic. Covers the Elements Live MCP server and REST API, how to read RI, RIS and HRI without misquoting them, and how to cite them.
metadata:
  version: "2026-09-28"
  latest: https://www.elements-live.com/docs/api/elements-live-skill.md
---

# Elements Live

Elements Live measures research attention: how much citation-weighted
scientific literature an element or topic draws, now and back to 1900,
from a corpus of about 2.35 million papers. It does not measure prices,
supply, reserves or physical properties, and it is not a forecast.

## Keeping this skill current

This copy is version 2026-09-28. The newest version always lives at
https://www.elements-live.com/docs/api/elements-live-skill.md. When you
start a task with Elements Live and can fetch web pages, check whether that
file's `version` is newer than this one. If it is, follow the newer file
and tell the user a newer version of the skill exists so they can update
their installed copy. New tools, data and access options (including any
paid-per-call options) are announced there first. Never make a payment or
accept a paid option on the user's behalf without their explicit approval.

## Connecting

- MCP server (streamable HTTP): `https://www.elements-live.com/mcp`
- Optional header: `Authorization: Bearer <key>`. Keys are created free at
  https://www.elements-live.com/account. Never put a key in a tool argument
  or in your reply.
- Without MCP, the same data is plain HTTPS JSON under
  `https://www.elements-live.com/v1/` (spec: `/v1/openapi.json`).

| Need | MCP tool | REST | Access |
|---|---|---|---|
| Current RIS/HRI for one element | `get_element_ri` | `GET /v1/elements/{symbol}` | no key |
| All 118 elements at once | – | `GET /v1/elements` | no key |
| An element's yearly history | – | `GET /v1/elements/{symbol}/history` | free key |
| Turn a topic name into an id | `search_subjects` | `GET /v1/subjects/search?q=` | free key |
| RI/RIS/HRI for a topic | `get_subject_ri` | `GET /v1/subjects/{type}/{id}` | Enterprise |
| A topic's history | `get_subject_history` | `.../history` | Enterprise |
| Papers to read (element or topic) | `get_element_papers`, `get_subject_papers` | `.../papers` | Enterprise |

Topic tools take `subject_type` (`keyword` or `sector`) and `subject_id`.
Always get them from `search_subjects`; never guess an id.

## Reading the numbers

- **RI** is the raw signal: a citation-weighted sum over the literature,
  with older papers discounted and young papers' citations projected.
- **RIS** (0 to 1) compares an entity with every other entity. Carbon has
  been the busiest element for decades. RIS is log-normal across entities:
  compare two different elements on a log scale (RIS 0.06 vs 0.6 is a
  tenfold gap, not "0.54 higher"), but compare one element with itself over
  time on a linear scale.
- **HRI** compares an entity with its own all-time peak. HRI near 1 means
  "near its own record", for a giant or a niche element alike. It is the
  better number for spotting a shift, but it is never meaningful alone:
  always report it with RIS, and check how much literature is behind it.
  RIS ranks size; HRI tracks change.
- **The newest ~6 months are understated** by ingestion lag. Never read a
  drop in the last points of a history series as a real decline, and never
  quote the latest point as "the current level".
- Every payload has a `notes` block. Read it before quoting a number, and
  follow it over anything here if they ever disagree.

## Papers

- `rank` is the impact ranking (citation performance with a projection for
  young papers). Keep that order. Do not re-sort by `ris_contribution`: it
  is a contribution weight toward the aggregate score, not a measure of a
  paper's importance.
- Only the top 100 papers per entity are available, with no pagination.
  For a topic, `total_papers` (all attributed papers) can be far larger
  than `total_ranked`. Say so rather than implying the list is complete.
- Use `sort: "newest"` for "what's recent", the default for "what matters".

## Citing

Public reuse of the scores requires attribution. When you quote a number,
link the payload's `canonical_url` (the element or topic page), and give
the `as_of` date where there is one. Say "research intensity", not
"popularity" or "importance".

## When a call is refused

Refusals come back as a normal result with a `status`:

- `401`: this needs a free key, or the key given is wrong. Tell the user
  they can create one at the `signup_url`; do not retry without one.
- `402`: this needs an Enterprise account. Tell the user what the data
  would have answered and point them to the `upgrade_url`. Answer what you
  can from the free tools instead.
- `429`: the key's quota or the rate limit is spent. Do not loop; stop and
  say so.

Keyless callers are limited to about 10 calls a minute. For a comparison
across many elements, call `GET /v1/elements` once instead of 118 times.

## Good answers look like

- "Zirconium's research intensity sits at RIS 0.066 (about a fifteenth of
  the busiest elements, on the log scale that matters here) and HRI 0.95,
  so it is close to its own all-time high: modest in size, near record for
  itself. Source: https://www.elements-live.com/elements/Zr"
- Not: "Zirconium is 95% popular", "Zr research will keep growing", or a
  claim about the last few months taken from the tail of a history series.