Excom MCP Server

Overview

EXCOM is an MCP (Model Context Protocol) server written in Elixir that aims to bridge the gap between AI assistants and external services through a lightweight, streaming‑friendly interface. The core problem it tackles is the friction developers face when trying to expose custom tools or data sources to AI agents: existing solutions often require complex API wrappers, manual serialization of state, and lack a standard protocol for tool invocation. By implementing MCP natively in Elixir, EXCOM provides a first‑class, type‑safe environment for building tool collections that can be consumed by Claude or other MCP‑aware assistants without additional plumbing.

At its heart, EXCOM offers a tool registry that lets developers register functions, resources, and prompts as declarative JSON schemas. These definitions are automatically served over HTTP, allowing the MCP client to discover available capabilities in a single request. The server also supports streaming execution, enabling long‑running operations to push incremental results back to the assistant in real time—a feature that is especially valuable for data‑intensive tasks such as live analytics or large file transformations. While the HTTP streaming support is still in early development, the architecture is designed to accommodate future enhancements like WebSocket or SSE endpoints.

Key capabilities include:

• Declarative tool definitions that enforce input/output contracts and provide rich metadata for the assistant’s UI.
• Resource handling for static or dynamic data sets that can be queried directly from the MCP client.
• Prompt templates that allow pre‑configured conversational snippets to be injected on demand.
• Sampling controls for adjusting the creativity or determinism of model outputs during tool invocation.

These features make EXCOM especially useful in scenarios where an AI assistant needs to interact with domain‑specific services—such as querying a company’s internal database, triggering CI/CD pipelines, or orchestrating microservices—without exposing the underlying infrastructure. Developers can compose complex workflows by chaining multiple EXCOM tools, each returning structured data that the assistant can consume and transform in subsequent steps.

Integration is straightforward for teams already using Elixir: the server exposes a clean HTTP interface that can be mounted behind any reverse proxy, and its JSON schemas are compatible with the MCP Inspector for rapid prototyping. Because EXCOM is built on top of Elixir’s robust concurrency model, it can handle high‑throughput workloads while maintaining low latency for streaming responses. These advantages position EXCOM as a compelling choice for developers who want a production‑ready MCP server that scales with their AI workloads.