AI Life Planner: A Local-First Personal Operating System

## The Goal Most productivity apps are either: - too rigid (you adapt to the tool), or - too unstructured (you become the database) AI Life Planner is an attempt to build a middle path: a local-first system with a consistent schema, where an agent can operate safely and quickly. ## What It Is AI Life Planner is a CLI-first personal operating system: - Projects, tasks, notes, and goals (PARA-inspired) - A database that keeps the structure honest - An agent interface (`ask` / `chat`) for natural language workflows - Integrations that are pragmatic about performance ## Integration Strategy: MCP *and* Gateways MCP is great when you want agent-native tool discovery and a standardized interface. But for high-frequency operations, startup overhead matters. The system supports daemon-style gateway paths for repeated calls. Here’s the “perception” framing I use when deciding:  ## System Shape At a high level: ``` planner CLI ├─ core models + database ├─ agents (NL → actions) └─ integrations ├─ MCP servers (where it fits) └─ CLI/daemon gateways (where speed matters) ``` This hybrid approach is what makes the system feel usable day-to-day: correctness and structure, without sacrificing responsiveness. ## Lessons Learned 1. **Schema is the safety rail**: it’s what makes agent behavior predictable. 2. **Local-first reduces friction**: it’s faster to adopt and easier to trust. 3. **Speed is a feature**: the same integration can feel “good” or “bad” purely based on startup overhead.