MCP Servers for OpenAI
Custom Model Context Protocol servers enabling modular AI workflow orchestration and reducing agent integration overhead across projects.
Overview
A suite of custom Model Context Protocol (MCP) servers that provide modular, plug-and-play tool access for OpenAI-powered agents. Built to reduce integration overhead across multiple agentic projects — instead of re-implementing tool connections per agent, each MCP server exposes a standardised interface that any compatible agent can call.
What is MCP?
MCP (Model Context Protocol) is Anthropic’s open standard for connecting AI models to external tools, APIs, and data sources. By wrapping business logic in MCP servers, agent integrations become hot-swappable — swap the underlying model or agent framework without rewriting tool logic.
Servers Built
- Web Search MCP — wraps Tavily and SerpAPI with query-normalisation and result ranking
- Database MCP — read/write access to PostgreSQL and SQLite with query sanitisation
- File System MCP — scoped read/write operations for document processing pipelines
- Calendar & Email MCP — Gmail and Google Calendar integration for scheduling agents
Technical Highlights
- FastAPI-based server implementations with OpenAI tool-call schema compatibility
- Shared authentication middleware across all servers
- Modular design: each server is independently deployable via Docker
- Reduced agent integration time by ~60% across downstream projects