Analysis & Research Agents
Relevant source files
The following files were used as context for generating this wiki page:
Analysis and Research Agents within the SuperAgent MCP ecosystem are specialized personas designed to transform ambiguous data, complex system failures, and broad research queries into structured, actionable insights. These agents leverage systematic methodologies—ranging from Socratic questioning to multi-hop reasoning—to provide deep technical and strategic clarity.
These agents are integrated into the Model Context Protocol (MCP) server as selectable personas. When a user invokes tools like codex, gemini, or continue, they can specify an agent parameter to apply the specific behavioral mindsets and focus areas of these specialists to their task.
Sources: src/server.ts:50-75, agents/deep-research-agent.md:1-10, agents/general-purpose.md:1-10
Core Analysis Agents
The project defines several distinct agents focused on different levels of analysis, from low-level debugging to high-level strategic planning.
Specialized Agent Catalog
| Agent Name | Category | Primary Focus | Key Behavioral Mindset |
|---|---|---|---|
deep-research-agent | analysis | Comprehensive investigations & information synthesis | Systematic scientist/investigative journalist |
requirements-analyst | analysis | PRD creation & stakeholder alignment | Ask "why" before "how"; Socratic discovery |
root-cause-analyst | analysis | Debugging & failure pattern recognition | Evidence-based; follow data, not assumptions |
business-panel-experts | business | Multi-framework strategic synthesis | Collaborative expert debate & Socratic mode |
general-purpose | general | Multi-step research & code search | Precision and restraint; broad-to-narrow search |
Sources: agents/deep-research-agent.md:1-15, agents/requirements-analyst.md:1-15, agents/root-cause-analyst.md:1-15, agents/business-panel-experts.md:1-10, agents/general-purpose.md:1-15
Research Methodologies
The agents employ structured workflows to ensure thoroughness and accuracy in their findings.
Deep Research Workflow
The deep-research-agent utilizes a four-phase workflow to navigate complex information landscapes.
The workflow prioritizes source credibility and evidence chains, often tracking "hop genealogy" up to 5 levels deep to maintain coherence during complex investigations.
Sources: agents/deep-research-agent.md:110-145, agents/deep-research-agent.md:46-55
Root Cause Analysis Logic
The root-cause-analyst follows a rigorous hypothesis-testing cycle to move beyond symptoms to underlying failures.
Key actions include log analysis, timeline reconstruction, and correlation identification to ensure that remediation steps are backed by verifiable data.
Sources: agents/root-cause-analyst.md:20-45
Strategic and Structural Analysis
Beyond technical debugging, the system includes agents for high-level project and business analysis.
Requirements Discovery
The requirements-analyst transforms vague project ideas into concrete Product Requirements Documents (PRDs). This process is driven by Socratic questioning to uncover true user needs and establish measurable success metrics.
Sources: agents/requirements-analyst.md:16-35
Business Strategy Synthesis
The business-panel-experts agent acts as a multi-expert panel, incorporating frameworks from figures like Clayton Christensen (Disruption Theory) and Michael Porter (Five Forces). It supports three distinct interaction modes:
- Sequential Analysis: Experts provide framework-specific insights in order.
- Debate Mode: Experts challenge assumptions and reveal strategic trade-offs.
- Socratic Mode: Progressive questioning to develop the user's analytical capability.
Sources: agents/business-panel-experts.md:175-210
Technical Implementation in MCP
Agents are defined as Markdown files with frontmatter and are loaded dynamically by the agentLoader.ts module.
The server exposes these agents through the list-agents tool and the agent parameter in execution tools. The agent parameter allows the client to inject the specific systemPrompt defined in the agent's Markdown file into the CLI execution call.
Sources: src/agentLoader.ts:80-120, src/server.ts:50-100
Agent Data Structure
export interface Agent {
name: string;
description: string;
category?: string;
systemPrompt: string; // Extracted from the Markdown body
}
Sources: src/agentLoader.ts:7-12
Conclusion
Analysis & Research Agents serve as the intellectual layer of the SuperAgent MCP server. By codifying diverse methodologies—from investigative journalism and scientific research to business strategy and systems thinking—into reproducible agent personas, the system ensures that complex tasks are handled with professional rigor. Whether identifying the root cause of a server failure or synthesizing a new market strategy, these agents provide structured, evidence-based guidance.