ffd-mcp
Parses SEC EDGAR Exhibit 107 XBRL fee disclosures from registration statements and returns structured fee data through the Model Context Protocol. No API keys required.
README
SEC EDGAR Filing Fee Disclosure MCP Server
MCP server for SEC EDGAR Exhibit 107 filing fee disclosure analysis.
Parses ffd: XBRL from registration statements (S-1, S-3, 424B, F-1, etc.) and returns structured fee data through the Model Context Protocol. No API keys required — SEC EDGAR is a free, public API.
7 Tools • 2 Resources • 1 Prompt • STDIO & Streamable HTTP
Tools
| Tool | Description |
|---|---|
get_fee_exhibit |
Fetch and parse a specific Exhibit 107 by accession number + CIK. Returns every security class, fee amounts, fee rate, and net fee due. |
search_fee_filings |
Find a company's recent registration filings that contain Exhibit 107 fee disclosures. Resolves tickers, names, or CIKs automatically. |
get_company_fee_history |
Fetch and parse all Exhibit 107 fee exhibits for a company in one call. Returns per-filing fee tables plus grand totals. |
compare_fee_rates |
Compare filing fee data across 2-8 companies side by side — latest form, total offering, net fee due, and fee rate. |
search_filings_by_fee |
Full-text search across all EDGAR filings since 1993, scoped to registration statements. Supports phrases, boolean operators, and wildcards. |
get_ffd_concepts_by_company |
Get a company's complete ffd: filing fee history by merging the EDGAR companyfacts API (fast, no lag — covers 424B/S-3/S-8) with Exhibit 107 document parsing (covers S-1/F-1 filings that companyfacts doesn't index). |
export_fee_data |
Query fee data and save as CSV. Three modes: fee_history (one row per filing), line_items (one row per security class), ffd_concepts (one row per XBRL fact). |
get_fee_exhibit
Parse a specific Exhibit 107 (EX-FILING FEES) from an SEC filing.
- Requires
accession_numberandcik - Optionally provide the
exhibit_filenamefor faster parsing, or let the server auto-detect it from the filing index - Returns all ffd: XBRL concepts: every security class, max aggregate offering, fee amounts, fee rate, and net fee due
- Includes retry logic for filings within the 24-48h EDGAR indexing lag window
search_fee_filings
Find a company's recent registration statement filings.
- Accepts ticker symbols (
RDDT), company names (Reddit), or CIK numbers - Defaults to all registration forms: S-1, S-3, S-4, F-1, F-3, F-4, F-6, 424B series, SC TO, PREM14A, DEFM14A, POS AM, S-8
- Returns accession numbers, form types, filing dates, and EDGAR URLs
- Surfaces a recency warning when results fall within the EDGAR indexing lag window
get_company_fee_history
Batch-parse all fee exhibits for a company across multiple filings.
- Parses up to 10 filings in a single call (configurable via
limit) - Returns the full structured fee table for each filing plus grand totals
- Filterable by form type and start date
compare_fee_rates
Side-by-side comparison of fee data for 2-8 companies.
- Shows the latest registration form, total offering, net fee due, and fee rate for each company
- Useful for benchmarking offering sizes across a peer set
search_filings_by_fee
Full-text search across all EDGAR filing documents, scoped to registration statements.
- Find companies that registered a specific security type, used a specific fee rule (e.g. Rule 457(o)), or mentioned specific terms
- Supports exact phrases, boolean operators (AND/OR/NOT), and wildcards
get_ffd_concepts_by_company
Complete fee history from two complementary data sources:
- Phase 1 — companyfacts API (fast, no indexing lag): Returns ffd: XBRL facts for 424B prospectus supplements, S-3/A, and S-8 employee plans
- Phase 2 — Exhibit 107 parsing (backfills the gap): Scans submissions for S-1/F-1 filings that the companyfacts API doesn't index, then fetches and parses each Exhibit 107
- Every entry is tagged with
source: "companyfacts"orsource: "exhibit107" - Optionally filter by specific ffd concepts (
FeeAmt,NetFeeAmt, etc.) or form types
export_fee_data
Query SEC fee data and save the results as a CSV file.
fee_historymode — one row per filing with summary totals: accession number, form type, filing date, total offering, total fee, net fee due, fee rate, security class countline_itemsmode — one row per security class across all filings: security type, class title, fee calculation rule, amount registered, price per unit, max aggregate offering, fee rate, fee amountffd_conceptsmode — one row per ffd: XBRL fact entry from the hybrid companyfacts + Exhibit 107 source: concept tag, label, value, unit, form, filing date, accession, source tag- Writes to
output_pathif provided, otherwise defaults to./<TICKER>_<mode>.csv - Filterable by form type and max filing count
Example prompts:
"Export Tesla's filing fee history as a CSV"
"Save Goldman Sachs' complete ffd data to C:\Users\me\Desktop\GS_fees.csv"
"Export the line-item breakdown for all of Reddit's S-1 filings"
CSV columns by mode:
| Mode | Columns |
|---|---|
fee_history |
accession_number, form_type, filing_date, entity_name, cik, exhibit_url, submission_type, total_offering, total_fee, previously_paid, fee_offset, net_fee_due, fee_rate, security_classes |
line_items |
accession_number, form_type, filing_date, entity_name, row_num, security_type, class_title, fee_calc_rule, amount_registered, price_per_unit, max_aggregate_offering, fee_rate, fee_amount, previously_paid |
ffd_concepts |
concept, label, value, unit, form, filed, accession, period_end, period_start, fiscal_year, fiscal_period, source |
Resources
| URI | Description |
|---|---|
ffd://fee-taxonomy |
Complete ffd: XBRL concept reference for Exhibit 107 — per-row and summary-level concepts, fee calculation rules, and historical SEC fee rates |
ffd://form-types |
Registration form types that require Exhibit 107 — S-1, S-3, F-1, 424B series, merger proxies, with key dates |
Prompts
| Prompt | Description |
|---|---|
fee_disclosure_analysis |
Guides structured analysis of a company's Exhibit 107 fee disclosures — identify the filing, parse the fee table, analyze the fee structure, and summarize |
Getting started
Prerequisites
- Node.js v20 or higher
Install from source
git clone https://github.com/davidtauriello/ffd-mcp.git
cd ffd-mcp
npm install
npm run build
Run locally (stdio)
EDGAR_USER_AGENT="YourApp you@email.com" npm start
Run locally (HTTP)
EDGAR_USER_AGENT="YourApp you@email.com" MCP_TRANSPORT_TYPE=http npm start
# Server listens at http://localhost:3010/mcp
# Health check at http://localhost:3010/health
Client configuration
Claude Desktop (stdio)
Add to claude_desktop_config.json:
{
"mcpServers": {
"filing-fee-mcp": {
"command": "node",
"args": ["/path/to/ffd-mcp/dist/index.js"],
"env": {
"EDGAR_USER_AGENT": "YourApp you@email.com"
}
}
}
}
Claude Code
claude mcp add filing-fee-mcp -- node /path/to/ffd-mcp/dist/index.js
Microsoft Copilot
Add to mcp.json:
{
"servers": {
"Filing Fee MCP": {
"type": "stdio",
"command": "node",
"args": [
"/path/to/ffd-mcp/dist/index.js"],
"env": {"EDGAR_USER_AGENT" : "YourApp you@email.com"}
}
},
"inputs": []
}
Set EDGAR_USER_AGENT in your environment.
Any MCP client (Streamable HTTP)
Point any client that supports Streamable HTTP at http://your-server:3010/mcp.
Docker
Build and run
docker build -t filing-fee-mcp .
docker run -d -p 3010:3010 -e EDGAR_USER_AGENT="YourApp you@email.com" filing-fee-mcp
The container defaults to HTTP transport on port 3010. Override the port with -e MCP_HTTP_PORT=8080 -p 8080:8080.
CSV export from Docker
The export_fee_data tool writes CSV files inside the container. Mount a volume to make them accessible on the host:
docker run -d -p 3010:3010 -e EDGAR_USER_AGENT="YourApp you@email.com" -v $(pwd)/exports:/app/exports filing-fee-mcp
Then use output_path: "/app/exports/GS_fees.csv" and the file appears in ./exports/ on the host.
Health check
curl http://localhost:3010/health
# {"status":"ok","server":"filing-fee-mcp-server","version":"1.0.0"}
Configuration
| Variable | Required | Default | Description |
|---|---|---|---|
EDGAR_USER_AGENT |
Yes | — | User-Agent header for SEC compliance. Format: "AppName contact@email.com". The SEC blocks requests without a valid User-Agent. |
EDGAR_RATE_LIMIT_RPS |
No | 10 |
Max requests/second to SEC APIs. Do not exceed 10. |
EDGAR_TICKER_CACHE_TTL |
No | 3600 |
Seconds to cache the company tickers lookup file. |
MCP_TRANSPORT_TYPE |
No | stdio |
Transport mode: stdio or http. |
MCP_HTTP_PORT |
No | 3010 |
HTTP server port (only used when transport is http). |
Project structure
src/
index.ts # Server entry point (McpServer + transport)
config/
server-config.ts # Environment variable parsing (Zod)
services/edgar/
edgar-client.ts # Rate-limited SEC EDGAR HTTP client
fee-exhibit-parser.ts # Exhibit 107 iXBRL/HTML parser
fee-exhibit-service.ts # Orchestrates client + parser
types.ts # Domain types
mcp-server/
tools/definitions/
get-fee-exhibit.ts # Parse a specific Exhibit 107
search-fee-filings.ts # Find registration filings
get-company-fee-history.ts # Batch-parse fee exhibits
compare-fee-rates.ts # Cross-company comparison
search-filings-by-fee.ts # Full-text search
get-ffd-concepts-by-company.ts # Hybrid companyfacts + parsing
export-fee-data.ts # CSV export
resources/definitions/
fee-taxonomy.ts # ffd: XBRL concept reference
form-types.ts # Registration form types
prompts/definitions/
fee-disclosure-analysis.ts # Guided analysis prompt
How it works
Exhibit 107
Since February 2022, the SEC requires all registration statement filers to include Exhibit 107 (EX-FILING FEES) — a machine-readable inline XBRL document disclosing every security class, offering amount, fee calculation rule, fee rate, and net fee due.
This server parses those exhibits using regex-based extraction of ffd: namespace XBRL facts from the iXBRL HTML. It handles both inline XBRL (<ix:nonNumeric>, <ix:nonFraction>) and standalone XML (<ffd:Tag>) formats, with a fallback HTML table parser for non-XBRL exhibits.
EDGAR indexing lag
EDGAR's submissions API lists new filings immediately, but the per-document index and full-text search can lag 24-48 hours behind acceptance. This server handles the gap with:
- Retry logic: When a document index returns empty for a filing within the 48h window, retries up to 3 times with increasing delays
- Recency warnings: Tools surface an advisory when results touch the lag window
- Companyfacts fast path: The
get_ffd_concepts_by_companytool reads from the companyfacts API (no lag) for 424B/S-3/S-8 filings, falling back to document parsing only for S-1/F-1 filings
License
Apache 2.0. See LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。