API Overview

DartLab’s current public Company flow is sections -> show(topic) -> trace(topic).

The core idea is not a list of parsers but a company map. First look at the company structure with sections, open the topic you need, and finally check whether the value came from docs, finance, or report.

Creating a Company

import dartlab

# DART
c = dartlab.Company("005930")
c = dartlab.Company("삼성전자")

# EDGAR
us = dartlab.Company("AAPL")

Current Public Surface

c.sections
c.show("BS")
c.trace("overview")

• c.sections: The public company board
• c.show(topic): Returns the topic payload
• c.trace(topic): Returns the source provenance

sections

sections is the canonical map of a company. It is a Polars DataFrame where each row is a disclosure block and period columns hold the original text.

c.sections                  # topic x period matrix
c.sections.periods()        # Available period list
c.sections.ordered()        # Ordered newest-first
c.topics                    # topic summary DataFrame (source, blocks, periods)

Key columns:

The intent of this board is:

• View a disclosure not as a vertical document but as a company map over a period axis
• Compare the same topic across multiple periods at a glance
• Share the same structure between AI, GUI, and Python

Use c.docs.sections when you need the pure docs source.

See the Sections Guide for detailed structure.

show(topic)

c.show("BS")
c.show("overview")
c.show("audit")

us.show("BS")
us.show("10-K::item1Business")

show(topic) opens a single topic.

The approximate source priority is:

• Financial figures: finance
• Structured disclosures: report
• Narrative / sections / original structure: docs

In other words, Company is not a raw source wrapper but a source-aware merged company object.

trace(topic)

c.trace("BS")
c.trace("dividend")
c.trace("overview")

trace(topic) explains which source was actually selected for the same topic.

Typical information checked here:

• Selected source
• Provenance
• Whether fallback was used
• Period coverage

diff()

diff() detects text changes between periods.

c.diff()                                # Change rate by topic
c.diff("businessOverview")              # Change history by period
c.diff("businessOverview", "2023", "2024")  # Line-by-line comparison

Years with high change rates may indicate business structure changes, new business entries, or additional risk factors.

Market-wide Financial Screening

Scan a single account or ratio across all listed companies.

import dartlab

# single account across all firms — wide DataFrame (rows=companies, columns=periods)
dartlab.scanAccount("매출액")                         # quarterly standalone revenue
dartlab.scanAccount("operating_profit", annual=True)  # annual basis
dartlab.scanAccount("total_assets", market="edgar")   # US EDGAR

# single ratio across all firms
dartlab.scanRatio("roe")                              # quarterly ROE
dartlab.scanRatio("debtRatio", annual=True)           # annual debt-to-equity

# available ratios (13: profitability, stability, growth, efficiency, cashflow)
dartlab.scanRatioList()

These are root-level functions only (not Company methods). Accepts both Korean names and English snakeIds. Reads 2,700+ parquet files in parallel via ThreadPool.

Source Namespace

When you need to go deeper, use the source namespace directly.

# DART
c.docs.sections
c.docs.retrievalBlocks
c.docs.contextSlices
c.finance.BS
c.report.audit

# EDGAR
us.docs.sections
us.finance.BS

OpenAPI

Separate from Company, public API wrappers are also provided.

from dartlab import OpenDart, OpenEdgar

d = OpenDart()
e = OpenEdgar()

e.company("AAPL")
e.filings("AAPL", forms=["10-K"])
e.companyFactsJson("AAPL")

The principle is:

• External interfaces are source-native
• Storage format is compatible with DartLab runtime
• The derived Company layer is a separate responsibility

MCP Server

DartLab includes a built-in MCP server that exposes 60 tools for Claude Desktop, Claude Code, and Cursor.

dartlab mcp              # start MCP server (stdio)
dartlab mcp --config claude-desktop   # show config for Claude Desktop

See the MCP Guide for setup instructions and the full tool list.

Stability

• DART core Company is centered on stable
• EDGAR is at a lower stability tier

See Stability Guide for detailed criteria.