daily-assistant
by asuranale
AI-powered daily task management MCP Server — recommend next task, inherit uncompleted todos, detect overdue, generate reviews. Deterministic ops in code (zero tokens), AI only when needed.
安装
claude skill add --url https://github.com/openclaw/skills文档
Daily Assistant MCP Server
A personal task management MCP Server. Deterministic operations (parsing, sorting, statistics) run in Python with zero token cost — AI only steps in when creativity is needed.
Setup
bash
git clone https://github.com/AsuraNale/daily-assistant-mcp.git
cd daily-assistant-mcp
python3 src/setup.py --auto # Windows: py src/setup.py --auto
The setup wizard creates a .venv, installs dependencies, sets up your data directory, and auto-configures your AI editor. No manual pip install or config editing needed.
Tools
| Tool | What it does |
|---|---|
recommend_next | Recommends the most important task to work on next |
get_today | Reads today's daily task file |
inherit_tasks | Carries over uncompleted tasks from yesterday |
check_overdue | Detects overdue task files |
generate_review | Generates end-of-day review with completion stats |
scan_split | Flags tasks >80min or missing time estimates |
Resources
| Resource | Content |
|---|---|
daily://today | Today's task file |
daily://dashboard | Dashboard overview |
daily://history | 7-day completion statistics |
Task Format
markdown
- [ ] Task description ⏱️45min 📅 2026-03-30 ⏫
Markers: ⏱️ = time estimate, 📅 = deadline, ⏫ = highest priority, 🔼 = high, 🔽 = low.
Design Philosophy
- Zero-token deterministic ops: Parsing, sorting, stats run in Python code
- AI only when needed: Task splitting, creative advice, context-aware suggestions
- Platform-independent: Windows, macOS, Linux. No Obsidian dependency
- Simple data: Plain Markdown files, edit anywhere