Earnings Call Behavioral Anomaly Detector

Detect when executives are hiding something. This MCP server analyzes earnings call transcripts and flags unusual behavioral patterns — speech complexity shifts, hedging spikes, evasion in Q&A, and linguistic drift over time.

Point any AI assistant at a stock ticker and get an instant behavioral risk assessment of the C-suite.

Two Ways To Use It

This Actor now supports both:

1. Direct MCP connection at /mcp
2. Standard Apify Actor runs using the input form or API

You can use either the dedicated MCP endpoint below or the Apify MCP wrapper form for this Actor's tools.

Supported Method: Direct MCP Connection

Connect your MCP client (Claude Desktop, Cursor, etc.) to:

https://muhdarifx--earnings-call-anomaly-detector.apify.actor/mcp

Header: Authorization: Bearer YOUR_APIFY_TOKEN

Then call the MCP tools normally via tools/list and tools/call.

Supported Method: Standard Actor Run

Run the Actor normally from Apify Console or API with a tool plus its arguments.

Example:

{
  "tool": "get_latest_anomalies",
  "ticker": "AAPL",
  "quarters_back": 8,
  "anomaly_threshold_sigma": 1.5
}

Example:

{
  "tool": "historical_drift",
  "ticker": "NVDA",
  "n_quarters": 4
}

Available Tools

Tool Name	Arguments
get_latest_anomalies	{ "ticker": "AAPL" }
compare_exec_baseline	{ "ticker": "AAPL", "exec_id": "timothy-d-cook", "quarter": "Q1 2026" }
get_evasion_signals	{ "ticker": "AAPL", "question_topic": "margins" }
historical_drift	{ "ticker": "AAPL", "n_quarters": 4 }

What You Get

4 MCP tools your AI assistant can call:

Tool	What it does
get_latest_anomalies	Full anomaly report for the most recent earnings call — risk scores, complexity shifts, hedging changes, Q&A evasion signals per executive
get_evasion_signals	Raw analyst-question / executive-answer pairs from the Q&A section, filterable by topic (margins, guidance, AI, etc.)
compare_exec_baseline	Compare one executive's speech patterns in a specific quarter against their historical baseline
historical_drift	Track how executive behavior has changed across multiple quarters — spot gradual shifts before they become news

Use Cases

• Investment research — Screen earnings calls for red flags before making decisions
• Due diligence — Compare executive behavior across quarters to spot deteriorating transparency
• Earnings preview — After a call drops, get an instant behavioral read on the management team
• Portfolio monitoring — Track behavioral drift across your holdings over time

Quick Start

Connect your AI assistant (Claude, GPT, etc.) to this Actor's MCP endpoint:

https://muhdarifx--earnings-call-anomaly-detector.apify.actor/mcp

Set the Authorization header to Bearer YOUR_APIFY_TOKEN.

Example MCP flow:

{ "jsonrpc": "2.0", "id": 1, "method": "initialize", "params": { "protocolVersion": "2025-03-26", "capabilities": {}, "clientInfo": { "name": "example-client", "version": "1.0.0" } } }

{ "jsonrpc": "2.0", "id": 2, "method": "tools/list", "params": {} }

{ "jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": { "name": "historical_drift", "arguments": { "ticker": "AAPL", "n_quarters": 4 } } }

Then just ask your AI:

"Analyze AAPL's latest earnings call for behavioral anomalies"

"Did TSLA's CEO dodge any questions about margins?"

"Compare NVDA's CFO speech patterns this quarter vs last 8 quarters"

"Show me behavioral drift for MSFT executives over the last 6 quarters"

What Gets Analyzed

For each executive on the call:

• Complexity shifts — Did their language suddenly get more or less complex vs their baseline? (Flesch-Kincaid grade level, σ deviation)
• Hedging patterns — Are they using more hedge words than usual? ("approximately", "potentially", "we believe")
• Prepared vs Q&A gap — Do they sound different when reading prepared remarks vs answering live questions?
• Q&A evasion — Raw question-answer pairs so your AI can judge if executives actually answered what was asked

Coverage

• Any US public company with earnings call transcripts
• Up to 20 quarters of historical data per ticker
• Automatic transcript sourcing and parsing
• Results cached per session for fast follow-up queries

Pricing

$0.50 per tool call — covers transcript scraping, parsing, and analysis. Follow-up calls on the same ticker use cached data at the same price.

Troubleshooting

• If your MCP client still shows Tool execution failed, reconnect the MCP server or refresh the tool registration so it picks up the latest deployed build and tool schema.
• If you want to bypass client-side caching entirely, call the dedicated /mcp endpoint directly or run the Actor normally with the tool input field.

Input Parameters

Parameter	Required	Default	Description
ticker	Yes	—	Stock ticker symbol (AAPL, TSLA, MSFT, etc.)
quarters_back	No	8	Number of quarters of history to analyze
anomaly_threshold_sigma	No	1.5	Sensitivity threshold — lower = more sensitive