Your AI investing agent
is only as smart as its data.

EDGAR
DART
EDINET
ESMA
A-share

With AkkruData or without —the gap in your agent's answers is plain to see.

Ask a professional question.Get a professional answer.

Deep questions, answered from the filings — not from memory. Every number can be audited and verified by anyone.

▾Agent: queryingAkkruDataMCP / API✓
Click any number to trace it to the source filing.
Company10-KLeases not commencedPurchase commitmentsUnrecog. tax
OracleFY26260.013.310.1
MetaFY25103.8131.016.5
AmazonFY2596.484.8+18.96.6
MicrosoftFY2592.7—24.7
AlphabetFY255.8+52.77.7+5.711.5
NvidiaFY2622.795.2+27.0+11.4+3.44.4
✓Agent: $634B of leases signed, none on a balance sheet yet. Oracle alone: 2.5× Meta.

Screen for exactly what you want.Never worry the numbers are wrong.

One call across the whole universe. Click any ratio to see the formula and the filing facts behind it.

▾Agent: queryingAkkruDataMCP / API✓
Click any number to trace it to the source filing.
TickerSectorROEDiv yieldFCF marginCap $bnPayout
APAMFinance
OPFIFinance
WUIndustrials
BKEConsumer Disc.
KTBConsumer Disc.
✓Agent: screened on ROE > 20%, dividend yield > 2.5%, FCF margin > 12%, market cap under $10B. 17 pass; these are the top 5 by ROE — small financials and cash-rich consumer names. I excluded H&R Block: its ROE is inflated by near-zero equity, not by earnings.

Backtests that can't cheat.

Every query takes an as_of_date. Ask for Nov 1, 2023 and you only get what was filed by then — no look-ahead bias, no peeking at restated numbers.

▾Agent: queryingAkkruDataMCP / API✓
The agent stamps every query with as_of_date = 2025-01-31. Amazon, one of the holdings:
as_of_date = 2025-01-31what the market could see
latest filing10-Q · Q3 FY2024 · filed 2024-11-01
FY2024 revenuenot filed yet
TTM revenue · thru 2024-09-30620,128← the backtest uses this
Amazon's FY24 10-K was filed Feb 7, 2025 — a week after the cutoff. Most APIs would hand it to your backtest anyway.
✓Agent: every pull in the backtest is cut off at its as-of date — Amazon's FY24 10-K, filed Feb 7, is invisible on Jan 31. Look-ahead bias: eliminated.

Research in bulk.

Companies, insiders, filings — compared side by side, from one question.

▾Agent: queryingAkkruDataMCP / API✓
Click any number to trace it to the source filing.
InsiderRoleShares sold% of holdingsUSD mTradesLast sale
Micron · 9 officers · 992,523 shares · $247.8M
Sanjay MehrotraCEO280,00042%96.81542026-05-29
Sumit SadanaCBO141,63839%32.552026-04-10
Mark MurphyCFO160,00042%32.1112025-10-30
Scott DeBoerCTO145,04749%25.342025-10-27
April ArnzenCPO100,00044%23.882026-04-01
Manish BhatiaEVP Ops106,62325%20.372026-01-22
Michael RayCLO36,60135%11.1492026-05-01
2 othersCAO, Sales22,61422%5.952026-04-14
Sandisk · 3 officers · 5,046 shares · $8.0M
Alper IlkbaharCTO2,0003.7%3.532026-06-01
Michael PokornyCAO2,4469.9%3.512026-05-12
Bernard ShekCLO6001.8%1.012026-06-03
David GoeckelerCEO——00no sales on file
✓Agent: Micron officers have been steady sellers — roughly 40% of their holdings, CEO included. Sandisk officers are holding; the CEO has not sold a share.

Built for autonomous research.

Ten years of filings don't fit any context window. On AkkruData your agent doesn't read — it queries, pulling exactly the facts each step needs.

Agent · autonomous run — deep dive on Micron, FY2016–FY2026
list_filingsMU · 2016–2026 · all form types
get_*_statementincome · balance sheet · cash flow · ×10 years
get_dimensional_breakdownrevenue by segment · by geography · ×10
get_filing_statementany footnote, on demand · tables, numbers, and the text as management wrote it
get_metrics_timeseries43 ratios × 40 quarters
list_insider_transactionsForm 3/4, every trade
list_filings8-K events, parsed
get_earnings_transcriptevery call, prepared remarks + Q&A · ×40 quarters
Reading the raw filings
Querying AkkruData
✓Agent: ten years of Micron — statements, segments, footnotes, insiders, events — researched, not crammed into a context window. Each step pulled exactly what it needed.
And a bigger window wouldn't fix it: the fuller the context, the more attention thins out and analysis quality drops. Querying beats reading.

What makes AkkruData different?

Statements as the company reported them. Not flattened into a template — every line, every subtotal, in the company's own structure.Show example

Footnotes included, in full. The tables, the numbers, and management's own wording — the detail most APIs throw away.Show example

Every number comes with its context. Unit, period, concept, and where it sits in the filing.Show example

Every metric shows its math. The formula and the inputs behind it.Show example

Every judgment call is disclosed. And you can override it.Show example

Ready to build the future of AI investing?