Mite Assistant MCP - AI Time Tracking via MCP

A personal weekend project born out of a real daily friction: having to context-switch between coding and filling in time entries in Mite. The idea was simple - let the AI assistant handle it while I stay in my editor. I built an MCP (Model Context Protocol) server that exposes Mite's REST API as a set of structured tools, so Claude Code or Codex can query, create, and update time entries through natural language.

Key Features & Details

• Protocol: Implements the Model Context Protocol (MCP) - the open standard for AI tool integration by Anthropic
• Auth: Stateless bearer token auth - the client passes the Mite API key with every request, no session storage needed
• Smart name resolution: Say "log 90 minutes to project Development" and the server resolves the name to a Mite project ID automatically
• Flexible time queries: Predefined frames (today, this_week, last_month, …) or exact date ranges in YYYY-MM-DD
• Full CRUD: list_time_entries, create_time_entry, update_time_entry, delete_time_entry + whoami
• Zero persistence: No database, no state - pure HTTP passthrough to the Mite API
• Health check: /health endpoint for uptime monitoring

graph LR
    A[Claude Code / Codex] -->|MCP over HTTP| B(mite-assistant-mcp)
    B -->|REST + API Key| C[Mite API]
    C -->|Time entries, Projects, Services| B
    B -->|Structured tool results| A

Why MCP? It gives AI assistants a typed, discoverable interface rather than raw HTTP calls - meaning the model understands what each tool does and can chain calls intelligently (e.g. look up a project ID before creating an entry).