MCP Scorecard

Mission StatementGitHub
← All posts

edgartools: The Highest-Scoring New Entry This Batch at 84

A Python library for SEC EDGAR data with 1,834 stars and perfect maintenance signals registers its own MCP server and scores 84 — the highest new entry this batch. Previously seen at score 71 under a different namespace, the project's direct registration demonstrates what clean metadata and a permissive license can do for a trust score.
io.github.dgunning

Two weeks ago, edgartools appeared in Registry Pulse 5 as the SEC Intelligence server under the tools.edgar namespace, scoring 71. That was a third-party MCP wrapper around the edgartools library. Now the project itself has registered a direct MCP server under its own namespace — io.github.dgunning/edgartools — and it scores 84. That makes it the highest-scoring new entry in this batch, and only the fourth server to crack the High Trust tier in the past three weeks.

What edgartools Is

edgartools is a Python library for accessing the SEC's EDGAR system — the Electronic Data Gathering, Analysis, and Retrieval system that houses every public company filing in the United States. Company 10-K and 10-Q filings, insider transaction reports (Form 4), institutional holdings (13F), proxy statements, 8-K material events, company facts, and financial statements are all accessible through a clean Python API.

Think of it as a free, open-source Bloomberg Terminal for SEC data. Financial analysts, quant researchers, fintech developers, and compliance teams use it to programmatically access the same filings that move markets. The library handles EDGAR's XML and XBRL parsing, converts financial statements into structured data, and provides search capabilities across the full corpus of SEC filings. 1,834 stars, 316 forks, 30 watchers. MIT license. Built and maintained by dgunning.

The MCP Integration

An AI agent connected to edgartools can now query SEC filings directly through natural language. The use cases are immediately practical:

  • “Show me Tesla's latest 10-K filing and summarize the risk factors.”
  • “What insider trades happened at Apple this week?”
  • “Pull the 13F holdings for Berkshire Hathaway's most recent quarter.”
  • “Compare revenue growth across Meta, Google, and Amazon for the last four quarters.”
  • “List all 8-K material event filings in the semiconductor industry from the past 30 days.”

This is the kind of research that currently requires a financial data terminal subscription or manual navigation of EDGAR's web interface. edgartools makes it an API call. The MCP server makes it a conversation.

The Score Breakdown

CategoryWeightAssessment
Provenance (30%)ExcellentSource repo present, MIT license (permissive), namespace matches owner, unique description
Maintenance (25%)Perfect52 active commit weeks, 7 contributors, 15 releases/year
Popularity (20%)Strong1,834 stars, 316 forks, 30 watchers
Permissions (25%)GoodMinor deductions for transport risk and credential sensitivity
Overall84

No flags. No secrets required. No red flags.

Why 84 Beats 71

The tools.edgar entry at score 71 and the io.github.dgunning/edgartools entry at score 84 both bridge the same underlying library to MCP. The 13-point difference comes down to metadata hygiene. The direct registration under the project owner's namespace (io.github.dgunning) provides a clean namespace-to-owner match, which the provenance category rewards. The MIT license is permissive rather than the more ambiguous licensing of a third-party wrapper. The direct entry benefits from the full weight of the source repository's signals — 52 active commit weeks, 7 contributors, 15 releases per year — without any indirection.

This is a useful lesson for library maintainers considering MCP: registering your own server under your own namespace, with your own repo as the source, will almost always outscore a third-party wrapper. The scoring model rewards direct provenance.

The Pattern Continues

edgartools joins a growing list of established projects choosing MCP as a distribution channel. Scrapling (score 92, 19.4k stars), Kubeshark (score 84, 11.8k stars), and now edgartools (score 84, 1.8k stars) all share the same profile: years of active development, existing user bases, and a decision to add MCP as an interface rather than building an MCP-native project from scratch. The star counts vary by an order of magnitude, but the trust scores converge — because the scoring model measures maintenance, licensing, and metadata quality, not just popularity.

For the financial data space specifically, edgartools at 84 sets a high bar. SEC data is one of the most valuable freely available datasets in the world, and having a High Trust MCP server that provides structured access to it is a meaningful addition to the registry.

Score: 84. No flags. MIT.

Sources: dgunning — GitHub · Scorecard: io.github.dgunning (score 84) · Previous entry: tools.edgar (score 71)

← Dev Tools Dominate: 56% of New MCP Servers Target DevelopersGearsystem: A Sega Master System Emulator in the MCP Registry →