Research Intensity (RI/RIS/HRI) for 118 elements and thousands of research topics, for developers and agents.
Create a free account, then generate a key from your account page. Send it as a bearer token:
curl -H "Authorization: Bearer elk_live_..." https://www.elements-live.com/v1/elements/Fe
Current element scores need no key at all; element history and topic search
are free with any key — this is the open, citable tier meant for indexing
and agent grounding. Topic/keyword data
(/v1/subjects/...) and paper lists need a Personal account,
and the Lab a Professional one, the same plans that show them on the
site; all of them can also be paid per call with
no account at all. History resolution follows your plan, as on the site:
yearly with any key, monthly from Personal, daily element history from Professional
(?resolution=). HRI is also a Personal feature: below Personal it
comes back as null, with a note saying where to get it.
Every response carries a canonical_url for attribution and a
notes block flagging how each metric should (and shouldn't) be read —
see methodology for the full explanation.
| Endpoint | With a key | Without a key | Returns |
|---|---|---|---|
| GET /v1/research?q= | Personal | $0.05 | Start here: communities + three reading lists for a question or paper (per page) |
| GET /v1/elements | No key | Free | All 118 elements |
| GET /v1/elements/<symbol> | No key | Free | One element |
| GET /v1/elements/<symbol>/history | Free | $0.05 | Full history (yearly; monthly on Personal, daily on Professional) |
| GET /v1/subjects/search?q= | Free | $0.01 | Resolve a query to a subject id |
| GET /v1/subjects/<type>/<id> | Personal | $0.01 | Topic/keyword RI/RIS/HRI |
| GET /v1/subjects/<type>/<id>/history | Personal | $0.10 | Full history (yearly or monthly) |
| GET /v1/subjects/<type>/<id>/elements | Personal | $0.10 | Elements most associated with the topic (top 10) |
| GET /v1/elements/<symbol>/papers | Personal | $0.02 (up to 20 papers), $0.10 (up to 100) | Top papers for an element |
| GET /v1/subjects/<type>/<id>/papers | Personal | $0.02 (up to 20 papers), $0.10 (up to 100) | Top papers for a topic |
GET /v1/research?q=<question>, or the MCP tool research
(listed first for agents), takes a question in plain words, or ?paper= a DOI
or title to start from. It returns the research communities and element sub-communities
the question is about, with why each matched and whether it is rising, and reading lists
in three lenses: foundational (the classics the question's own papers cite most, found by
citation, not keywords),
now (highest impact, young papers' citations projected) and
newest (the last two years). Each paper has a DOI and a one-line reason.
With a key it needs a Personal account.
$0.05 per page without a key.
?page_size= sets papers per lens (1–10, default 10);
?lenses=foundational,now picks lenses.next gives page 2's URL and price
before you ask for it; pages run to 5.unmatched_terms. A question nothing matches
is a 404 and never charged.
Describe the papers you want with a filter and get back how many match, how they
spread over the years and over research communities, and a binding price for each
deliverable. GET /v1/lab/quote?q=..., POST /v1/lab/quote with
{"query": ...} or {"markdown": ...} (the filter in a
```query block), or the MCP tool lab_quote.
With a key, the Lab needs a Professional account.
$0.01 per quote without a key.
+Li +Co year:2015..2026 (Nb OR Ta) +superconduct* +"lithium niobate" -O
+term or a bare term is required; -term excludes;
(a OR b) accepts either; "two words" is one exact topic
name; word* expands over topic names (up to 25);
year:2015..2026, year:2020.. or year:2024.ignored, never silently dropped.broader.quote_id. Above 2,000 matches, years and communities come from a fixed
sample (flagged); above 100,000 matches the packet is a licence, not a price.GET /v1/lab/run?quote_id=...&depth=papers_20 (or
papers_100), POST /v1/lab/run, or the MCP tool
lab_run returns the top papers matching the quoted query, ranked by impact.
It costs the depth's quoted price less the $0.01 the quote cost. An expired, edited or
stale quote (the data updated since) is refused before any payment is asked.
Agents and scripts without a key can pay for each call in USDC on Base or Solana
using x402. Call any priced endpoint
without a key: the 402 response's PAYMENT-REQUIRED header
lists the price and how to pay. Sign the payment with an x402 client and repeat the
request with a PAYMENT-SIGNATURE header; the receipt comes back in
PAYMENT-RESPONSE. On the MCP server the same works inside the tool call:
the unpaid call returns the payment terms, and the paid call carries the payment in
params._meta["x402/payment"].
x-payment-info.lab_quote $0.01, lab_run the quoted price of the depth, less the $0.01 quote credit, search_subjects $0.01, get_subject_ri $0.01, get_subject_history $0.10, get_subject_papers $0.02, get_element_papers $0.02, research $0.05.
The paper endpoints return title, DOI, venue, year, citations and a link —
enough to ground an answer or build a reading list, plus the curation behind each
paper's place: impact_score (the ranking score, with young papers'
citations projected), citation_percentile and cluster. Accepts
?sort=impact|newest|foundational (foundational = most-cited first), ?limit= (default 20, max 100), and on the
element route ?keyword=. Three things to know:
total_ranked is how many are reachable here;
total_papers (topics only) is how many the topic really has. If you
need the whole corpus, that's a bulk export.ris_contribution. It's the paper's
contribution weight toward that entity's aggregate score, not a measure of the
paper's importance. rank is the real ordering — citation impact
with a projection for papers too recent to have accrued citations yet.Full machine-readable reference: /v1/openapi.json.
Add https://www.elements-live.com/mcp as a remote MCP connector with your key
in the Authorization header. Tools: get_element_ri (no key needed),
search_subjects (free with a key), and get_subject_ri, get_subject_history,
get_subject_papers, get_element_papers (Personal, or
paid per call without a key).
The paper tools are the grounding ones — use them when the question is
what to read, not how active a topic is.
A ready-made skill teaches your AI assistant which tool to call, how to read RI, RIS and
HRI without misquoting them, and how to cite them.
Download elements-live-skill.zip and upload it
in Claude's skill settings, or drop the folder into
~/.claude/skills/ for Claude Code. The skill checks
the latest version itself and tells you when
there is a newer one.
Enterprise API access, or the full corpus rather than per-call access? Contact us.