Integration MCP Servers
Relevant source files
The following files were used as context for generating this wiki page:
Integration Model Context Protocol (MCP) servers within the eng-platform provide specialized connectors and orchestration layers that bridge AI agents with external data sources, research tools, and engineering metrics. These servers are designed to be domain-agnostic, solving universal challenges such as context drift, multi-step research orchestration, and database introspection across various environments including PostgreSQL, MySQL, and cloud storage providers like Dropbox.
Sources: mcp-servers/README.md, README.md:14-25
Core Research and Analytics Architecture
The platform utilizes a sophisticated Adaptive Research Engine (ARE) and Query Refinement Engine (QRE) system. This architecture orchestrates multiple specialized servers to handle complex research requests, ensuring high confidence through statistical analysis and pattern detection.
ARE/QRE System Flow
The following diagram illustrates the multi-layered orchestration of research requests through the core integration servers.
Sources: mcp-servers/README.md
Integration Server Catalog
The eng-platform categorizes its integration servers based on their functional role in the engineering lifecycle.
Storage and Database Connectors
These servers provide standardized interfaces for AI agents to interact with structured and unstructured data repositories.
| Server | Port | Key Capabilities | Use Cases |
|---|---|---|---|
| postgres | 9384 | Schema introspection, sanitized query execution, connection pooling. | Database migrations, data analysis. |
| mysql | 9385 | Schema introspection and SQL execution for MySQL/MariaDB. | Generic database access. |
| dropbox | 9387 | File upload/download, directory sync, shared link management. | Automated backup, report distribution. |
| ragflow | 3010 | Hybrid search (keyword + semantic), multi-KB federation. | Internal knowledge base search. |
Sources: mcp-servers/README.md, audits/legalai-system/phase1-audit.json
Specialized Research and Analytics Servers
These servers handle the logic of processing information and measuring the quality of AI outputs.
1. Basin Analyzer (Port 9383)
The Basin Analyzer detects "context drift" by measuring the stability of LLM outputs. It runs queries multiple times (N samples) to identify "basins of attraction"—stable response patterns.
- Metrics: Epsilon (ε) indicates distance between basins (lower = more confident), while Basin Count (n) identifies distinct patterns.
- Query Classification: Categorizes queries as FACTUAL, INTERPRETIVE, or COUNTERFACTUAL based on response stability.
Sources: mcp-servers/README.md, CHANGELOG.md:179-188
2. Team Analytics (Port 9385)
Provides proactive engineer health monitoring by analyzing git history and team memory contributions.
- Health Scoring: Aggregates metrics for Code Quality (35%), Knowledge Sharing (25%), Velocity (25%), and Collaboration (15%).
- Proactive Alerts: Generates Critical, Warning, or Positive alerts based on score thresholds (e.g., Critical if score < 50).
Sources: mcp-servers/team-analytics/README.md
Configuration and Deployment
Integration servers are typically configured via a .mcp.json file. The platform supports multiple transport layers, including HTTP and Server-Sent Events (SSE).
Setup Sequence for Established Projects
When adding these servers to an existing repository, the system follows a non-destructive integration sequence.
Shared Environment Variables
Many integration servers require specific API keys for semantic embeddings and LLM processing:
VOYAGE_API_KEY: Required for semantic embeddings in the Basin Analyzer.DEEPSEEK_API_KEY: Required for LLM calls within research orchestration.MEMORY_FILE_PATH: Used bysmart-memoryandteam-analyticsto point to.shared/team-memory.json.
Sources: PROMPT-ADD-TO-EXISTING.md, mcp-servers/README-SETUP.md
Summary
The Integration MCP Servers in eng-platform provide a robust framework for extending AI capabilities into data management and research validation. By utilizing the ARE/QRE architecture and specialized tools like the Basin Analyzer and Team Analytics, the platform ensures that AI-driven operations remain stable, measurable, and highly integrated with the existing engineering stack.
Sources: mcp-servers/README.md, README.md:130-141