Extensibility

MCP servers and extensibility

Connect external MCP servers to extend Claude's tool set with database access, API integrations, cloud management, and custom capabilities.

Updated 2026-03-21

What is MCP

The Model Context Protocol (MCP) is an open standard for connecting AI models to external tools and data sources. MCP servers expose capabilities — database queries, API calls, file operations, or any custom logic — that Claude can call during a conversation.

Onevium includes a built-in MCP server for schedule management and supports connecting any number of external MCP servers.

Add an MCP server

Go to Settings > MCP Servers and click Add Server. Each server needs:

  • Name — A unique identifier (e.g., "postgres", "github", "slack").
  • Command — The executable that starts the server (e.g., "npx", "uvx", "node").
  • Arguments — Command-line arguments for the server process.
  • Environment variables — Key-value pairs for API keys, database URLs, and other secrets.

Server scoping

MCP servers can be configured at two levels:

  • User-level (global) — Available in all projects. Stored in ~/.claude/settings.json.
  • Project-level — Only available when working in a specific directory. Stored in {project}/.claude/settings.json.

Built-in Onevium MCP server

The Onevium MCP server runs in-process and is always available. It exposes schedule management tools that Claude can call during any conversation: create_schedule, list_schedules, get_schedule, update_schedule, and delete_schedule.

Popular MCP server examples

The MCP ecosystem has servers for common integrations:

  • Database — PostgreSQL, MySQL, SQLite for direct query access.
  • Version control — GitHub, GitLab for issue and PR management.
  • Communication — Slack, Linear, Notion for team tool integration.
  • Cloud — AWS, GCP, Cloudflare for infrastructure operations.
  • Custom — Build your own in any language that supports the MCP protocol.