SEC Compliance Data: Where to Find It and How to Use It

SEC compliance data is technically free. It's all public, sitting on EDGAR, available to anyone with a browser. In practice, getting from "available" to "usable" is where most teams burn weeks.

We've spent two years building pipelines that turn EDGAR filings into structured data — for DealCharts, for our outcome monitoring system, for client implementations. The raw material is the same for everyone. The difference is the pipeline that turns it into something a compliance team — or an AI agent — can actually work with.

Here's a map of what's out there, where it lives, and how to make it useful.

What SEC Compliance Data Actually Covers

The SEC's EDGAR system contains filings from roughly 150,000 entities — public companies, investment funds, asset-backed trusts, broker-dealers, investment advisers. There are about 16,000 registered investment advisers alone.

For compliance purposes, the filings that matter most break into a few categories:

Corporate Filings

10-K — annual reports. Financial statements, risk factors, MD&A, executive compensation. The most comprehensive disclosure a public company makes.

10-Q — quarterly reports. Same structure as 10-K but for interim periods. Less comprehensive but more frequent.

8-K — material events. Triggered by specific events: executive changes, acquisitions, earnings, material agreements. These are time-sensitive — they have to be filed within four business days of the event.

DEF 14A — proxy statements. Board composition, executive compensation, shareholder proposals. Required before annual meetings.

Fund Filings

NPORT-P — quarterly portfolio holdings. Every registered investment company (mutual funds, ETFs) files complete position-level holdings — every security, every CUSIP, every dollar value. This is the filing that DealCharts parses for its fund holdings data.

N-CEN — annual fund census data. Registration details, service providers, fee structures.

Structured Finance

ABS-EE — asset-backed securities. Loan-level data for securitized pools — auto loans, mortgages, credit cards, CLOs. Filed at issuance and updated monthly or quarterly.

Insider Activity

Forms 3, 4, 5 — insider ownership and transactions. Required when officers, directors, or 10%+ shareholders buy or sell company stock.

Where the Data Lives

All of this is on EDGAR, but EDGAR isn't one system — it's sort of a collection of systems built over 30 years. Knowing which endpoint to use for what saves a lot of time.

EDGAR Full-Text Search

The EFTS API lets you search across all filings by keyword. Useful for finding specific disclosures or entities by name. Rate-limited to 10 requests per second.

CompanyFacts API

The CompanyFacts API returns structured XBRL data for any company by CIK. Financial statement line items, standardized, with dates. This is the best endpoint for quantitative fundamental data.

Filing Index

The filing index lets you browse all filings for a given entity by type and date range. Returns links to the actual filing documents.

XBRL Viewer

The SEC's inline XBRL viewer renders financial data from filings with tagged line items. Useful for human review but not great for automation.

Bulk Data

For large-scale processing, EDGAR provides bulk download archives organized by quarter. These are the raw filing packages — you'll need to parse the XML/XBRL yourself.

The Problem: Available Isn't Usable

All of this data is accessible. Very little of it is ready to use.

NPORT-P filings are XML documents buried inside filing archives. To get fund holdings, you need to download the archive, find the right XML document, parse it, map the fields, and normalize the output. Per filing. For thousands of funds. Every quarter.

10-K and 10-Q filings contain XBRL-tagged financial data — but the tagging is inconsistent across companies, fiscal years, and even between different sections of the same filing. Getting consistent time-series data out of XBRL is a real engineering problem.

8-K filings are semi-structured at best. The event types are standardized (Item 1.01, Item 2.01, etc.) but the content is free-text. Extracting structured information requires either manual review or NLP — and NLP without provenance just creates another trust problem.

This is the gap we've been filling. DealCharts parses NPORT-P filings into structured JSON — fund identifiers, holdings with CUSIPs and dollar values, asset categories, filing dates. Every data point links back to the source filing by accession number. The search API and facts endpoints give you the parsed data directly — no key required, CC-BY 4.0 license.

For CMBS and ABS deals, same pattern — loan-level data parsed from ABS-EE filings, organized by deal, traceable to source.

Building Compliance Monitoring on SEC Data

Once the data is structured, monitoring becomes a pipeline problem. Here's the architecture we use:

Detect

Poll EDGAR for new filings. The ATOM feeds update in near-real-time. When a new 10-K, 8-K, or NPORT-P appears for a tracked entity, the pipeline picks it up.

Parse

Extract structured data from the filing. For XBRL filings, map the tagged data to canonical fields. For XML (NPORT-P, ABS-EE), parse the document structure. For free-text (8-K items), extract the structured elements and flag the rest for review.

Compare

This is where the value is. Compare the new filing against the prior state. What changed? Did a company's risk factors expand? Did a fund's top holdings shift? Did a BDC's portfolio concentration increase?

The comparison requires temporal versioning — you need to know what the system believed before the new filing arrived. Without that, you can detect new filings but you can't detect changes.

Alert with Provenance

Every alert links back to the source filing. Not "Company X filed a 10-K" but "Company X filed 10-K on 2026-03-10 (accession 0001234567-26-000789), compared against prior filing from 2025-03-11 — here's what changed, here's a link to both documents."

That's an alert with provenance. It's auditable. An examiner can verify it. An AI agent can cite it.

What This Looks Like in Practice

Our SEC Filing Monitor runs this pipeline for any set of companies. You pick your issuers, choose your filing types, select a delivery channel (email, Slack, webhook), and the system monitors EDGAR continuously.

Every notification includes a link to the source filing, a summary of what changed, and a provenance record that captures the full detection chain — when the filing appeared, when it was parsed, what was compared, and what was delivered.

For fund holdings, DealCharts runs the same pipeline at scale — thousands of NPORT-P filings parsed into structured JSON, searchable by fund name, queryable by API. The data is used by portfolio managers, data vendors, and AI systems that need fund holdings with provenance.

For AI Systems

If you're building AI applications on SEC data, the compliance data layer is the foundation. An LLM can summarize a filing. A RAG pipeline can retrieve relevant disclosures. An agent can monitor for changes.

But all of those capabilities require structured, traceable data underneath. The model is only as trustworthy as its inputs — and if the inputs don't carry provenance, the outputs can't be audited.

This is why we built the SEC EDGAR MCP Server — a Model Context Protocol server that lets Claude, Cursor, or any MCP client query EDGAR filings in natural language. The structured data layer does the heavy lifting. The AI system gets clean inputs with source attribution. The compliance team gets outputs they can verify.

SEC compliance data is free. Making it trustworthy is the hard part. That's where we spend our time.

---

Related:

• SEC Filing Monitor — Automated EDGAR monitoring with provenance
• SEC EDGAR MCP Server — Query filings from Claude, Cursor, or any MCP client
• SEC EDGAR API Guide — CompanyFacts endpoint, CIK lookup, and Python examples
• How to Look Up ETF Holdings — Fund holdings from NPORT-P filings via API
• Is Data Lineage Required for Compliance? — The broader regulatory picture