io.github.EvidInvest/aether-developer MCP Server

This MCP server provides cited search over SEC filings, earnings transcripts, and EU regulation, using a fetch mode. It exposes 14 tools to connect Aether to an LLM, an IDE, or custom code. Results are formatted as ready-to-cite evidence payloads, including the exact filed passage and links to sec.gov and EUR-Lex.

🛠️ Key Features

Cited search over SEC filings, earnings transcripts, and EU financial regulation
Fetch mode
14 tools
Evidence payloads with exact filed passage, accession-numbered citation, sec.gov/EUR-Lex link, and a confidence score

🚀 Use Cases

Agent or RAG pipelines that require citations from SEC and EU sources
Workflows needing evidence-first outputs rather than uncited answers

⚡ Developer Benefits

Integration with LLMs, IDEs, or own code
Structured, ready-to-cite outputs (passage + accession citation + source links)
Connection guides and REST API reference included in docs

⚠️ Limitations

Described as evidence/passage-focused: without filing-backed results, it does not provide answers.

aether-developer

Everything you need to connect Aether — the search engine built for agents, over SEC filings, earnings-call transcripts, and EU financial regulation — to your LLM, your IDE, or your own code.

Every result is a ready-to-cite evidence payload: the exact filed passage, an accession-numbered citation, the sec.gov / EUR-Lex link, and a confidence score. No filing, no answer — your agent stops inventing numbers.

code

aether-developer/
├── docs/                — connection guides + REST API reference
│   ├── chatgpt.md       — ChatGPT Developer Mode setup
│   ├── mcp.md           — Claude / Cursor / Cline / any MCP client
│   ├── cursor.md        — Cursor specifics (one-click install)
│   └── search.md        — REST API reference (/v1/tools/*), the issuer contract, fetch mode
├── skills/              — Agent Skills (drop into ~/.claude/skills)
├── mcp/                 — @evidinvest/aether-mcp (stdio MCP server, npm)
├── clients/
│   ├── typescript/      — @evidinvest/aether-sdk (npm)
│   └── python/          — evidinvest-aether-sdk (PyPI)
└── examples/            — runnable demos + config snippets

Connect your LLM (no code)

MCP endpoint: https://api.aether.evidinvest.com/mcp · Auth: OAuth (sign in when your client prompts — free account).

Client	How	Guide
ChatGPT	Settings → Apps & Connectors → Developer mode → Create connector with the URL above	`docs/chatgpt.md`
Claude (web/Desktop)	Settings → Connectors → Add custom connector with the URL above	`docs/mcp.md`
Claude Code	`claude mcp add --transport http aether https://api.aether.evidinvest.com/mcp`	`docs/mcp.md`
Cursor		`docs/cursor.md`
Cline / Continue / other stdio MCP	`npx -y @evidinvest/aether-mcp` in the client's MCP config	`docs/mcp.md`

Then ask: "What does Apple's latest 10-K say about supply-chain risk? Quote the passage with the sec.gov link." Name the company in the question — every tool takes an issuer, and naming it is what makes the answer scoped rather than a relevance-ranked sweep of every filer.

Teach your agent the craft (skills)

MCP gives the agent the tools; the skills/ directory teaches it the craft — tool routing, citation discipline, point-in-time transcript retrieval:

bash

cp -r skills/aether-research ~/.claude/skills/   # Claude Code, personal

See skills/README.md for Claude Desktop / other frameworks.

Call the API from code

Four REST endpoints — the same tools MCP exposes (full reference):

code

POST /v1/tools/search   ← start here (unified)
POST /v1/tools/financial_search    POST /v1/tools/transcript_search    POST /v1/tools/regulation_search

Name the company, then narrow. Pass issuer on every call and the answer is scoped to that filer: fast and precise. Omit it and the call searches every issuer at once — slower, relevance-ranked only, and the response carries a quality_caveat. scope: "issuer" with no caveat is the only guarantee you got a filtered answer, so read both fields.

And query is optional: omit it, pass an identifier plus form_type / fiscal_year / section, and you get fetch mode — that filing's sections in filing order, no ranking, mode: "fetch", tens of milliseconds. If you already know which document you want, don't rank it.

TypeScript / Node

bash

pnpm add @evidinvest/aether-sdk

import { AetherClient } from "@evidinvest/aether-sdk";

const aether = new AetherClient({ apiKey: process.env.AETHER_API_KEY });

// search: the question in `query`, the company in `issuer`
const out = await aether.financialSearch({
  query: "supply-chain concentration risk",
  issuer: { ticker: "AAPL" },
  fiscal_year: 2025,
  limit: 5,
});
if (out.quality_caveat) console.warn(out.scope, out.quality_caveat);
for (const c of out.results) {
  const start = c.metadata?.retrieval?.body_offset ?? 0; // skip the EDGAR cover page
  console.log(c.citation, c.metadata?.source_url, c.text.slice(start, start + 200));
}

// fetch: no query — Apple's latest 10-K risk factors, in filing order
const risks = await aether.financialSearch({
  issuer: { ticker: "AAPL" },
  form_type: ["10-K"],
  section: "Item 1A",
  limit: 5,
});
console.log(risks.mode); // "fetch" — and no `confidence` on any hit

Python

bash

pip install evidinvest-aether-sdk

python

from aether import AetherClient

with AetherClient(api_key="ak_...") as aether:
    out = aether.financial_search(
        query="supply-chain concentration risk",
        issuer={"ticker": "AAPL"},
        fiscal_year=2025,
        limit=5,
    )
    if out.quality_caveat:
        print(out.scope, out.quality_caveat)
    for c in out.results:
        start = c.body_offset or 0          # skip the EDGAR cover page
        print(c.citation, c.metadata.get("source_url"), c.text[start : start + 200])

    # fetch: no query — Apple's latest 10-K risk factors, in filing order
    risks = aether.financial_search(
        issuer={"ticker": "AAPL"}, form_type=["10-K"], section="Item 1A", limit=5
    )
    print(risks.mode)   # "fetch" — and `confidence` is None on every hit

curl / any language

bash

curl -sS https://api.aether.evidinvest.com/v1/tools/search \
  -H "authorization: Bearer $AETHER_API_KEY" \
  -H "content-type: application/json" \
  -d '{"query":"supply chain risk Taiwan","issuer":{"ticker":"TSM"},"limit":5}' | jq .

Docs

docs/search.md — REST API reference: all three tools, params, response shape.
docs/regulation_search.md — EU financial-regulation search deep-dive (29 acts), filters, curl.
docs/mcp.md — connect any MCP client (remote URL or stdio), env overrides.
docs/chatgpt.md — ChatGPT Developer Mode walkthrough.
docs/cursor.md — Cursor one-click install.

Get an API key

https://aether.evidinvest.com/developer/keys — anonymous calls work but are rate-limited; keys are free.

What's in the corpus

SEC filings — 10-K / 10-Q / 8-K, registration statements, prospectuses, press exhibits; ~10 years of S&P 500 and beyond.
Non-US registries — Sweden (Bolagsverket), Japan (EDINET), Korea (DART) annual reports. Reach a non-US filer by name — issuer: {company_name: "Sivers Semiconductors"} — or scope a whole market with jurisdiction: ["SE"|"JP"|"KR"].
Earnings calls — speaker-attributed transcripts + furnished press exhibits, with point-in-time filters (order: "earliest" finds first mentions) and the issuer's own fiscal_year + quarter.
EU regulation — MiFID II, MiCA, DORA, the AML package; article-level citations.

3 months free. Verify your email when you sign up and you're on the trial tier automatically — up to 5,000 API calls/hour on Aether's search tools, free for 3 months. No card required; it reverts to the free tier (100 calls/hour + paid credits) when the trial ends.

Contribute a client library

The TypeScript and Python clients are intentionally tiny — one bearer-token fetch wrapper + typed shapes for the /v1/tools/* endpoints. Porting to Go, Rust, Java, etc. is ~150 lines. Open a PR under clients/<lang>/. Whatever you write, surface scope, quality_caveat, mode and each hit's body_offset: a client that drops them leaves its users unable to tell a precise answer from an approximate one. Response fixtures to test against live in clients/fixtures/.

License

Apache-2.0 — see LICENSE.

query	string	optional	Natural language search query. OPTIONAL — omit it (or pass an empty string) to select FETCH MODE, which ranks nothing and returns rows in filing order. Fetch mode REQUIRES an identifier: pass `issuer` (ticker, cik or company_name) or a single-element `tickers`, optionally narrowed with `fiscal_year`; a query-less call with no identifier is rejected. Fetch fans out to sec/jp/kr/transcripts only — EU regulation has no issuer, so "every article, newest first" answers nothing.
issuer	object	optional	The company the question is about. Pass at least one of ticker (e.g. "NVDA"), cik (e.g. "1045810"), or company_name (required for non-US issuers, which have no US ticker). Forwarded to every corpus that can use it. Strongly recommended on every call.
fiscal_year	any	optional	Issuer's fiscal year as the company labels it (NVIDIA FY2027 = year ending Jan 2027). Forwarded verbatim to every corpus: on filings it is resolved from XBRL DEI fiscal-year focus where available, else from the issuer's fiscal year-end; on earnings calls it is that corpus's own labelled fiscal year. An empty year widens ±1 with a caveat.
scope	string	optional	issuer (default): every corpus is scoped to the named company. cross_company: search all issuers — for questions like "who names NVIDIA as a supplier". Cross-company results are slower, relevance-ranked only, and always carry `quality_caveat`.
tickers	array	optional	Legacy alias of `issuer`, still accepted. One symbol (e.g. ["IBM"]) is read as issuer.ticker; several (e.g. ["IBM", "MU"]) is a cross-company question and is labelled scope=cross_company, though the filter to those companies still applies. Prefer `issuer`.
limit	integer	optional	Max hits to return (1-50, default 10).
corpora	array	optional	Optional corpus filter. Omit to search everything (recommended). Values: sec, jp, kr, regulation, transcripts. An explicit value always wins: a filing filter (form_type / accession_number / section) narrows an OMITTED fan-out to sec/jp/kr, but naming corpora here overrides that.
form_type	array	optional	Only return these filing forms, for example ["20-F"] or ["6-K"]. Forwarded to the filing corpora (sec/jp/kr). Because an earnings segment has no form type and an EU article has none either, setting this ALSO narrows an omitted `corpora` to sec/jp/kr rather than merging rows the filter could not reach.
accession_number	string	optional	Exact SEC accession number. When set, the filing corpora are hard-scoped to that filing, and an omitted `corpora` narrows to sec/jp/kr. It also counts as the identifier fetch mode requires, so `{accession_number}` with no `query` is a legal fetch.
section	any	optional	Filing section(s), matched case-insensitively as a substring of the stored label — "Item 1A" matches "Item 1A. Risk Factors". In fetch mode this is how you ask for one part of a filing. Filing corpora only; it narrows an omitted `corpora` to sec/jp/kr.
quarter	any	optional	Fiscal quarter(s) — "Q2" or ["Q1","Q2"]. Exact, case-insensitive. Applies to the EARNINGS-CALL corpus only (with issuer.ticker and fiscal_year it names one call, the natural fetch-mode request); the filing and regulation corpora ignore it.
return_format	string	optional	Response shape for the filing corpora. "chunk" (the unified default) returns the matching window only — it keeps a five-corpus merge small. "section" returns the full SEC section the chunk belongs to, "both" returns both. The earnings-call and regulation corpora ignore it; their unit is a speaker turn / an article.

query	string	optional	Natural language search query. OMIT IT for fetch mode: with no query nothing is embedded or scored and the matching sections come back in filing-date-desc then document order. Fetch mode requires an identifier (issuer.ticker / issuer.cik / issuer.company_name / cik[] / accession_number) — a query-less call without one is rejected.
issuer	object	optional	The company the question is about. Pass at least one of ticker (e.g. "NVDA"), cik (e.g. "1045810"), or company_name (required for non-US issuers, which have no US ticker). Scopes retrieval to that filer: fast and precise. Strongly recommended on every call, and REQUIRED in fetch mode.
fiscal_year	any	optional	Issuer's fiscal year as the company labels it (NVIDIA FY2027 = year ending Jan 2027). Resolved from XBRL DEI fiscal-year focus where available, else from the issuer's fiscal year-end; an empty year widens ±1 with a caveat.
scope	string	optional	issuer (default): results limited to the resolved company. cross_company: search all filers — for questions like "who names NVIDIA as a supplier". Cross-company results are slower, relevance-ranked only, and always carry `quality_caveat`.
domain	string	optional	public_equity = single-company filing search; supply_chain = cross-company supplier/customer relationship evidence; auto (default) = detect from the query.
limit	number	optional
fields	array	optional
profile	string	optional	Retrieval profile. `hybrid` (default) ranks by the engine's own fused score (saturated lexical + vector cosine), highest first. `bm25` is the lexical-only fallback. `hybrid_rerank` and `hybrid_rerank_tickerprior` are accepted as aliases of `hybrid` for one release — the cross-encoder they named was removed on 2026-09-11.
return_format	string	optional	Response shape. 'section' (default) returns the full SEC section the matching chunk belongs to. 'chunk' returns the matching window only. 'both' returns both.
accession_number	string	optional	Exact SEC accession number. When set, retrieval is hard-scoped to that filing.
cik	array	optional	Exact SEC CIK values to scope retrieval. Values are normalized to 10 digits.
form_type	array	optional	Only return these filing forms, for example ["20-F"] or ["6-K"].
section	any	optional	Filing section(s), matched case-insensitively as a substring of the stored label — "Item 1A" matches "Item 1A. Risk Factors". In fetch mode this is how you ask for one part of a filing.
exclude_form_type	array	optional	Exclude filing forms. Company searches already suppress fund/proxy forms by default.
prefer_recent	boolean	optional	Prefer newer filings when relevance scores are close. Automatically enabled for latest/current/guidance queries.
jurisdiction	array	optional	Scope to non-US issuer jurisdictions (ISO-3166 alpha-2): "SE" (Sweden/Bolagsverket), "JP" (Japan/EDINET), "KR" (Korea/DART). These carry no US ticker and are otherwise buried by SEC-centric ranking. US filings have an empty jurisdiction tag, so this filter is for non-US scoping only — omit it to include US (the default). A named non-US issuer surfaces even without this filter, via the company-name anchor.

query	string	optional	Natural language query. OMIT IT for fetch mode: with no query nothing is embedded or scored and the matching segments come back newest call first, then in the order they were spoken. Fetch mode requires a ticker.
issuer	object	optional	The company whose calls you mean. Pass `ticker` (e.g. "NVDA"); this corpus is keyed on ticker, so that is the field that scopes it. Strongly recommended on every call.
fiscal_year	any	optional	Fiscal year(s) the question concerns — the ISSUER'S OWN fiscal year, as printed on its cover page, matched exactly against each segment's fiscal-year label. Offset-fiscal-year issuers work as stated: NVDA fiscal_year 2027 returns the calls held during calendar 2026, AAPL 2025 the Q1-Q4 FY2025 calls. If the requested year matches nothing the search retries once at ±1 and says so in `quality_caveat`. Independent of date_from/date_to, which bound the CALL DATE and still apply on top.
scope	string	optional	issuer (default): results limited to the issuer's own calls. cross_company: search every issuer's calls — slower, relevance-ranked only, and always carries `quality_caveat`. Setting it drops the ticker filter, so a caveated answer is never secretly a scoped one.
ticker	string	optional	Optional ticker filter (e.g. NVDA). Legacy alias of `issuer.ticker`.
lookback_quarters	number	optional	Keep calls within last N quarters (default: no filter)
speaker_role	string	optional	Optional speaker role filter: CEO / CFO / Analyst / Operator
date_from	string	optional	Point-in-time lower bound on call date, inclusive ISO YYYY-MM-DD
date_to	string	optional	Point-in-time upper bound on call date, inclusive ISO YYYY-MM-DD
order	string	optional	Sort order. 'earliest'/'latest' sort matched hits chronologically (use 'earliest' for onset/first-mention queries); 'relevance' sorts by rank score.
source_type	string	optional	Provenance filter: asr_call = our ASR of the real call (Q&A); furnished_transcript = issuer transcript exhibit; press_release = 8-K press exhibit.
quarter	any	optional	Fiscal quarter(s) — "Q2" or ["Q1","Q2"]. Exact, case-insensitive. With issuer.ticker and fiscal_year this names one call, which is the natural fetch-mode request.
limit	number	optional
profile	string	optional

query	string	required	Natural language query
celex	string \| array	optional	The act(s) to anchor on, by CELEX number — e.g. 32014L0065 (MiFID II), 32023R1114 (MiCA), 32013R0575 (CRR), 32022R2554 (DORA), 32016R0679 (GDPR), 32024R1624 (AMLR). This is the subject of the call: pass it whenever you know which act you mean. Omitting it searches the whole 29-act corpus as scope=cross_company, with a quality_caveat.
scope	string	optional	issuer (default, requires `celex`): results come from that act only. cross_company: the whole corpus — slower, relevance-ranked only, always carries `quality_caveat`. A `celex` filter you pass always applies.
doc_type	string	optional	Optional filter: regulation \| directive \| rts \| its \| decision
article	string	optional	Optional single-article filter, e.g. "20" or "12a"
chunk_type	string	optional	Optional filter: paragraph \| article_intro \| recital \| table \| annex
aml_topics	string \| array	optional	Optional topic tag filter: cdd \| edd \| pep \| str_reporting \| governance \| reporting \| transaction_monitoring
prefer_consolidated	boolean	optional	Prefer EUR-Lex consolidated text over original OJ text at equal relevance (hybrid only). Leave off to favor the as-published OJ text.
limit	number	optional
profile	string	optional

ticker	string	optional	Ticker, e.g. NVDA. Resolved to a CUSIP internally.
cusip	string	optional	9-character CUSIP, e.g. 67066G104. Use when the ticker is unmapped.
issuer	object	optional	The company the question is about, in the same shape financial_search and search take — so an agent can carry ONE issuer object across tools. Alias only: `issuer.ticker` fills in `ticker` when you have not set it. Ownership is a structured lookup, so there is no scope or quality_caveat here — `coverage_caveat` is this tool's equivalent.
period	string	optional	Quarter end as YYYY-MM-DD, e.g. 2026-03-31. Defaults to the most recent loaded quarter.
limit	number	optional	Holders to return, 1-100 (default 20), ranked by value.

io.github.EvidInvest/aether-developer

io.github.EvidInvest/aether-developer MCP Server

🛠️ Key Features

🚀 Use Cases

⚡ Developer Benefits

⚠️ Limitations

Topics

search

financial_search

transcript_search

regulation_search

holdings_by_security

holdings_by_manager

list_partners

partner_search

partner_proxy_search

seller_signup

seller_publish_document

seller_register_endpoint

seller_list_my_documents

seller_list_my_endpoints

aether-developer

Connect your LLM (no code)

Teach your agent the craft (skills)

Call the API from code

TypeScript / Node

Python

curl / any language

Docs

Get an API key

What's in the corpus

Contribute a client library

License

Install

Configuration

Related servers

manager	string	optional	Manager name, partial match, e.g. 'Bridgewater' or 'BlackRock'. Largest matching filer wins.
cik	string	optional	Exact 10-digit filer CIK, e.g. 0002012383. Preferred when known.
issuer	object	optional	The company the question is about, in the same shape financial_search and search take — so an agent can carry ONE issuer object across tools. Alias only: `issuer.cik` fills in `cik` and `issuer.company_name` fills in `manager` when you have not set them. NOTE this tool identifies the FILING MANAGER, not the issuer held — for who owns a stock, use holdings_by_security. Ownership is a structured lookup, so there is no scope or quality_caveat here — `coverage_caveat` is this tool's equivalent.
period	string	optional	Quarter end as YYYY-MM-DD. Defaults to the most recent loaded quarter.
limit	number	optional	Positions to return, 1-200 (default 25), ranked by value.

slug	string	required	Lowercase kebab-case identifier shown in attribution.
org_name	string	required
contact_email	string	required
description	string	optional	Free-text description shown to agents in list_partners.
invite_code	string	optional	Optional operator-issued invite. Bypasses per-IP rate limit and auto-approves.

seller_api_key	string	required	Your seller API key (aether_sk_…).
external_doc_id	string	required
title	string	optional
doc_type	string	optional
license	string	optional
ticker_coverage	array	optional
content	string	required	Document text. Up to ~2 MB.
source_url	string	optional
metadata	object	optional