MCP Guide

Disclosure Alpha ships two Model Context Protocol servers for AI agents in Cursor, Claude Desktop, and other MCP hosts.

Prerequisites: pip install "disclosure-alpha[mcp]" (see Installation) and SEC_USER_AGENT for ticker-based tools.

Choose a bundle

Entry point

Best for

Tools

disclosure-alpha-mcp-analyst

Ticker discovery + scoring

2 tools + taxonomy resource

disclosure-alpha-mcp-builder

Raw HTML pipeline control

5 low-level tools

Legacy disclosure-alpha-mcp aliases to the analyst bundle.

Analyst bundle

Start the server (stdio transport — configure in your MCP host):

export SEC_USER_AGENT="YourName your@email.com"
disclosure-alpha-mcp-analyst

Tools:

Tool

Description

list_company_filings_tool

List 10-K / 10-Q filings for a ticker and fiscal year

score_company_filing_tool

Score a filing by ticker, fiscal year, form type, optional quarter; optional scoring_model_version (default v2)

Resource: disclosure://taxonomy/v1 — component weights and version strings.

Builder bundle

disclosure-alpha-mcp-builder

Tools:

Tool

Description

extract_sections_tool

Extract sections from filing HTML

compute_section_metrics_tool_wrapper

Metrics + diffs from section JSON

diff_sections_tool

Diff two section texts

score_deterministic_tool_wrapper

Aggregate scores from metrics JSON; optional scoring_model_version (default v2)

score_filing_html_tool_wrapper

Full pipeline on HTML (optional prior HTML); optional scoring_model_version (default v2)

Use the builder bundle when your agent already has filing HTML and needs step-by-step pipeline access.

Scoring model: all scoring tools default to deterministic_scoring_v2. Pass scoring_model_version=deterministic_scoring_v1 for the legacy scale. See Versioning and Reproducibility.

Cursor configuration example

Add to your MCP settings (paths may vary):

{
  "mcpServers": {
    "disclosure-alpha": {
      "command": "disclosure-alpha-mcp-analyst",
      "env": {
        "SEC_USER_AGENT": "YourName your@email.com"
      }
    }
  }
}