Finance Toolkit
500+ transparent financial methods calculated from raw statements, not third-party endpoints.
제작자가 쓴 설명(영어 원문)
Rust MCP server for SEC EDGAR data — company filings, financial concepts, and company info
이 세 가지는 절대 하나의 점수로 합치지 않아요. 의견, 측정, 인기는 서로 다른 것이니까요.
이 서버는 아직 쉬운 말로 된 요약을 쓰지 못했어요. 아래는 레지스트리에서 읽어 온 사실이에요.
쓰고 있는 AI 앱을 고른 다음 설정을 복사하세요.
제작자가 레지스트리에 설치 정보를 올리지 않았어요. 아래 제작자의 README를 확인하세요.
내 컴퓨터에서 실행돼요. 설치하면 AI 앱에 도구 목록이 나타나요.
누구나 볼 수 있는 정보를 점검했고, 코드는 읽지 않았어요. 해당하지 않는 항목은 개수에서 뺐어요. · 2026년 11월 7일 점검
레지스트리 이름이 GitHub 계정 brechanbech 기준으로 확인되었고, 이 계정이 저장소를 소유하고 있어요
아직 GitHub에서 이 저장소를 읽어 오지 않았어요
아직 GitHub에서 이 저장소를 읽어 오지 않았어요
내 컴퓨터에서 실행되는 서버라 인터넷 엔드포인트가 없어요
내 컴퓨터에서 실행되며, 대신 로컬 설정을 사용해요
내 컴퓨터에서 실행돼요. 도구 목록은 인터넷 엔드포인트가 있는 서버에서만 읽을 수 있어요
레지스트리에서 이 카테고리의 활성 서버 전체와 비교했어요. 스타는 게시자가 저장소를 소유한 경우에만 세요.
해당하는 기본 점검 1개 중 1개를 통과했어요. 코드는 읽지 않았어요. 키나 토큰을 넣기 전에 서버가 어떤 권한을 요구하는지 확인하고, 꼭 필요한 권한만 있는 키를 쓰세요.
레지스트리에 필수 키가 나와 있지 않아요. 다만 처음 쓸 때 로그인을 요구하는 서버도 있어요.
내 컴퓨터에서 실행돼요. AI 앱이 필요할 때 실행하고, 데이터는 제작자의 서버를 거치지 않아요(서버 자체가 온라인 서비스를 호출하는 경우는 제외).
GitHub 저장소는 마지막으로 – 변경됐고, 최신 버전(0.10.0)은 11일 전 레지스트리에 올라왔어요.
Claude Desktop, Claude Code, Cursor, VS Code처럼 MCP를 지원하는 앱에서 쓸 수 있어요. 내 컴퓨터에 Rust도 있어야 해요.
같은 카테고리에는: Finance Toolkit, SolidInvoice, Korea Tax Calculator
같은 카테고리에서 다른 제작자가 만든 서버예요. 스타 순으로 정렬했어요.
카드 보는 법:★ GitHub 스타i세 기둥: 리뷰 · 점검 · 사용 정도i업데이트 = 최신 버전을 레지스트리에 공개한 날i
500+ transparent financial methods calculated from raw statements, not third-party endpoints.
Create and send invoices and quotes, manage clients, and track payments in SolidInvoice.
Korean tax calculation and tax-audit judgment MCP: 96 tools with statutory citations.
제작자의 저장소에서 쓰인 그대로 가져왔어요. PopNx MCP가 검토하거나 보증한 내용이 아니에요. HTML, 이미지, 스크립트는 제거했어요. 링크를 누르면 이 사이트를 벗어나요.
A Model Context Protocol server that gives Claude Desktop access to SEC EDGAR data — company filings, financial statements, and company info. No API key needed. It runs as a single native binary, with no Node, Python or Docker runtime behind it.
EDGAR is a set of public REST endpoints with no API key, and the documentation at sec.gov/search-filings/edgar-application-programming-interfaces is reasonable. So strictly speaking, an MCP isn't needed. What this wrapper buys you:
company_tickers.json and parse it from scratch. The MCP caches it for an hour.CY{period} prefix on the XBRL frames endpoint, the trailing I for instant concepts, the zero-padded CIK, the dash-stripped accession number in archive URLs. Easy to get wrong on the fly; encoded once in the tool.companyfacts responses are tens of MB and filings.recent is laid out as column-major parallel arrays. Both are usable, but a language model would burn a lot of context reformatting them. The MCP returns trimmed, row-shaped JSON.If you only pull one or two filings a year, you don't need this — just give Claude the URL. At any higher volume, the wrapper pays for itself by turning "explain the URL grammar in every conversation" into "ask a natural question."
Install with Cargo (Rust 1.88 or newer):
cargo install sec-mcp
This compiles the rustls crypto backend (ring), which has a small amount of C, so you need a C compiler on the build host — but no cmake or assembler. macOS (Xcode Command Line Tools) and most desktop Linux distros already have one; a minimal Linux image may not — install it first, e.g. apt install build-essential on Debian/Ubuntu.
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": { "sec-edgar": { "command": "sec-mcp", "args": [] } } }
Restart Claude Desktop.
On your first SEC-related request Claude will explain that a contact email is required by the SEC EDGAR fair-access policy and ask your permission before proceeding. Your email is stored locally in ~/Library/Application Support/sec-mcp/config.toml and is only used in the HTTP User-Agent header sent to the SEC — it is not shared anywhere else.
| Tool | Description |
|---|---|
sec_configure | First-run setup — saves your contact email |
sec_lookup_cik | Resolve a ticker to its SEC CIK number, or search every EDGAR filer by name: fund managers, private issuers and individuals with no ticker included |
sec_company_info | SIC code, industry, state of incorporation, fiscal year end, addresses |
sec_recent_filings | Recent filings (10-K, 10-Q, 8-K, etc.) with direct URLs — pages into older history automatically when a form-type filter needs it. With no ticker, the latest filings across all of EDGAR in real time, by form, with each filing's parties and an 8-K's items |
sec_filing_text | The text of a filing document or exhibit (10-K, 10-Q, 8-K, earnings release EX-99.1, proxy): its section index, any section by item, a search within it, or pages from any point. Given a filing, lists its documents by type |
sec_full_text_search | Search the body text of filings since 2001, exhibits included — the only tool here that finds companies by what they *said*, rather than needing the company up front. Filter by form, ticker and date (amendments excluded unless asked for); every hit carries a direct document URL. With no query, lists filings by form and date instead, for any period since 2001 |
sec_insider_transactions | Forms 3, 4 and 5 — who bought or sold, when, how many shares and at what price, with transaction codes in plain language and footnotes resolved. Also Form 144, an insider's notice of a proposed sale, with any 10b5-1 plan date |
sec_institutional_holdings | Form 13F — what an institutional manager (Berkshire, Bridgewater, BlackRock) holds each quarter, as positions with value, share of portfolio and voting authority, filterable by issuer. Issuers are named as filed, since 13F has no tickers |
sec_private_offerings | Form D — money raised without a registered offering by private companies, funds and public companies: amount offered and sold, investors, the exemption claimed, brokers paid, and the officers, directors and promoters named |
sec_beneficial_ownership | Schedules 13D and 13G — who owns more than 5% of a company, with voting and dispositive power, percent of class and, for a 13D, the holder's stated purpose. Given a holder's CIK, what that holder owns above 5% |
sec_financial_concept | Historical financial data from XBRL (revenue, net income, EPS, assets…) — latest 20 facts by default, paging back into older history |
sec_list_tickers | Search/list all SEC-registered tickers with exchange info — 50 per page, with paging |
sec_company_facts | Discover all XBRL concepts (metrics) a company reports |
sec_xbrl_frames | Cross-company comparison of a financial metric for a given period |
Wherever a tool takes a ticker, it also takes a CIK (1163302, 0001163302 or CIK0001163302). SEC's ticker list holds only current tickers, so a company that was acquired, delisted or renamed is reachable only by CIK. sec_lookup_cik with name finds it, and full-text search hits and frames rows carry one.
10-K/A and friends)A form ending in /A amends one filed earlier. A 10-K/A is often not a complete annual report. Rule 12b-15 requires only the full text of each item being changed, so the most common amendment is Part III alone (directors, pay, ownership), filed when the proxy statement misses its 120-day window. Others add or fix exhibits. The rest are restatements that replace the original's figures. Apple's January 2010 10-K/A, for example, refiled fiscal 2009 under new revenue-recognition rules.
sec-mcp therefore excludes amendments unless you ask for them:
sec_financial_concept's annual/quarterly period filters match base forms only. The amended period is already covered by the original filing, and admitting both would report two facts for one period. Pass include_amendments: true to admit them when the restatement is what you want: Apple's fiscal 2009 revenue is $36.5B on the original 10-K and $42.9B on the 10-K/A. amended_facts then counts them.sec_full_text_search with a form filter drops amendments. EDGAR itself does not: it filters on a filing's *root* form, so forms=10-K also returns 10-K/A. To get amendments, request the form directly (10-K/A), list both (10-K,10-K/A), or pass include_amendments: true. Any amendment returned is marked amendment: true.Because that search filtering happens after EDGAR's, two parts of the response behave differently from a plain search:
total_is_upper_bound, or total_is_approximate when it has also saturated at 10000. Each page reports amendments_excluded.next_offset. A page can return fewer than limit hits because amendments were skipped, so offset + returned would revisit them.sec_financial_concept| Concept | What it is |
|---|---|
Revenues | Total revenues |
NetIncomeLoss | Net income / loss |
EarningsPerShareBasic | Basic EPS |
EarningsPerShareDiluted | Diluted EPS |
Assets | Total assets |
Liabilities | Total liabilities |
StockholdersEquity | Shareholders' equity |
OperatingIncomeLoss | Operating income / loss |
CashAndCashEquivalentsAtCarryingValue | Cash and equivalents |
CommonStockSharesOutstanding | Shares outstanding |
cargo test # offline; hits no network
There is also an opt-in live suite that drives the built server over stdio against the real SEC EDGAR APIs — the thing offline tests can't see, like EDGAR changing a response shape or a URL-construction quirk slipping through. It is skipped unless SEC_MCP_LIVE_EMAIL is set, which both opts in and supplies the contact email the fair-access policy requires (the address is read from the environment, never committed):
SEC_MCP_LIVE_EMAIL="you@example.com" cargo test --test live_smoke -- --nocapture
The suite runs sequentially through one server process, so it stays well within EDGAR's rate limit.
MIT — see LICENSE.md for details.
Ownership-verification token for the MCP registry (read from this crate's rendered README on crates.io):
Registry ownership token:
mcp-name: io.github.brechanbech/sec-mcp
A version is immutable once published — crates.io yanks, it does not delete, and a number is never reused — so the order matters:
cargo test and cargo clippy --all-targets clean.version in `Cargo.toml`.version in `server.json` — *both* fields, the top-level one and the one under packages[]. Nothing enforces this: no script, no CI, no test reads the file, and it is the step that goes missing. This manifest stayed correct, but the registry *entry* did not: 0.5.0, 0.5.1 and 0.5.2 all shipped to crates.io while the registry still advertised 0.4.4, because step 6 was never run.cargo publish --dry-run, which compiles *from the packaged copy* and so catches a crate depending on a file it does not ship. Then cargo publish. Give both --target-dir target/package-verify. Without it, the verify build shares target/debug with ordinary builds under the same unit hash, so the binary's앞부분만 보여 드려요. 나머지는 저장소에서 읽어 보세요