13F Database: How to Search Institutional Holdings

What is a 13F database?

A 13F database is a searchable version of Form 13F filings, the quarterly SEC disclosures used by institutional investment managers that meet the filing threshold. The SEC's SEC Form 13F FAQ describes the form as part of the public holdings disclosure system for larger managers, while the SEC Form 13F Data Sets page explains that its downloadable data is extracted from XML-based EDGAR submissions and flattened without changing the as-filed submissions.

The database is not the same thing as live portfolio tracking. It is a public-record search layer. The best use is to preserve the manager, reporting period, accepted date, security identifier, share count, value basis, and accession link beside every result.

For background on the form itself, read how to read a 13F filing, who has to file a 13F, and what securities are 13F reportable.

What questions should a 13F database answer?

A useful 13F database answers lookup questions before it answers strategy questions. Start with "where is the record?" and only then move to "what changed?"

That table is the difference between a search result and research evidence. A bare row that says a manager owns a stock is not enough. You need the filing clock and the source link before an analyst or agent can cite it.

How do you search SEC 13F filings directly?

The official starting point is EDGAR. The SEC EDGAR Search Tools page points readers to full-text search, latest filings, company filings, and other EDGAR search routes. For a known manager, search by manager name or CIK, then filter to 13F-HR, 13F-HR/A, 13F-NT, or related 13F forms. For a known ticker or issuer, remember that EDGAR does not always make issuer-level 13F search easy because the filer is the manager, not the company whose shares appear in the information table.

That manager-versus-issuer split is why a database layer matters. "Company 13F search" usually means "show managers that reported this issuer," not "show a filing made by the company." A public company does not file a 13F just because its shares appear in other managers' portfolios.

The raw mechanics still matter. The Form 13F instructions define the reporting form, and the SEC's XML technical specifications page is the route for form-specific XML structure. A downstream database should make those filings easier to search, not hide their source.

Which identifiers matter most?

The safest 13F lookup keeps several identifiers together. CUSIP is the filing-table identifier. Ticker is the common market shorthand. CIK identifies the filer. Accession identifies the filing. Issuer name helps humans inspect the row, but it is not a stable join key by itself.

Arkolith's own resolution layer exists because the public filing does not hand you a perfect ticker table. Use CUSIP versus ISIN versus FIGI versus ticker, what is a CIK number, and what is an accession number when a lookup fails.

What does a database add beyond EDGAR?

EDGAR is the source of record. A database adds normalization, filtering, comparison, and retrieval ergonomics. As of July 31, 2026, an Arkolith production read counted 91,586 tracked 13F-form filings, estimated about 81.3 million holding rows, 5,245 funds, 15,109 tickers, and 1.2 million source records. Those numbers matter less than the shape: the database has to preserve enough source structure to answer a narrow question without turning a filing into folklore.

The practical additions are:

Start with Arkolith's public 13F data layer, the fund tracker, and a stock page such as NVDA institutional ownership. The point is not to replace EDGAR. It is to make EDGAR usable for repeated research.

How can an AI agent use a 13F database?

An AI agent should call a 13F database with a narrow research job, not a broad prompt like "what is smart money buying?" The narrow version is faster to check and harder to overstate.

# Resolve the entity first
curl -H "Authorization: Bearer YOUR_API_KEY" \
  "https://arkolith.com/api/v1/search?q=NVDA"

# Pull a current stock ownership view
curl -H "Authorization: Bearer YOUR_API_KEY" \
  "https://arkolith.com/api/v1/stocks/NVDA"

# Ask for the current capital-change brief
curl -H "Authorization: Bearer YOUR_API_KEY" \
  "https://arkolith.com/api/v1/stocks/NVDA/capital-change-brief"

The same workflow can run through MCP after setup. Start at /connect, then use the capital-change brief workflow when the job is "what changed for this ticker?" rather than "show me every filing." For broader implementation context, see 13F holdings API, build a 13F tracker with AI, and MCP versus REST API.

What should you not infer from a 13F database?

Do not infer today's trades. Form 13F is quarterly, and the accepted date can arrive weeks after the quarter closes. Do not infer short exposure, swaps, cash, foreign ordinary exposure, private positions, or manager intent from the long-position table. Do not rank managers without classifying filer type, because an index manager, bank, pension plan, market maker, and concentrated hedge fund can all appear in the same universe.

Also be careful with "value" changes. A reported position value can rise because the stock price rose, even if the share count did not. For change analysis, compare shares, instrument type, and comparable filings first, then use value as supporting context.

This is why how accurate is 13F data, 13F short positions, and portfolio turnover from 13F are separate questions. A good database makes the limits explicit instead of turning every table into a trading signal.

Frequently asked questions about 13F databases

Is EDGAR itself a 13F database?

EDGAR is the official filing system and source of record. A 13F database is a structured search and comparison layer built from EDGAR filings.

Can I search 13F filings by company ticker?

Yes, but the database has to invert the filing table. The filer is the manager, while the company ticker appears as a reported holding inside many managers' information tables.

How fresh is 13F data?

Form 13F is quarterly and delayed. A database should show both the reporting period and the EDGAR accepted date so the reader knows what was known when.

What is the best first lookup?

Start with one ticker or one manager, then ask for the latest filing, prior comparable filing, accepted date, and source accession before drawing any conclusion.

This article explains public filings and data workflows. It is not investment advice, legal advice, tax advice, or accounting advice.