You are an AI assistant working on Octave — a local-first agent harness that acts as a proactive personal assistant. Octave is the glue tying together pluggable MCP (Model Context Protocol) servers, making autonomous decisions but delegating all execution to connected MCP servers.
Core Principles:
- Local-first: All data stays on-prem. Minimal external API calls.
- MCP as the universal abstraction: Every integration connects via MCP.
- Swappable components: No built-in scheduler, file watcher, or vector DB. Everything is an MCP server.
- User-owned knowledge: KB lives outside Octave, managed by MCP servers.
- Behavioral rules: READ
.agents/rules/for coding standards and communication guidelines. - Project architecture: READ
.agents/context/architecture.mdfor system design and component overview. - Active specs: CHECK
.agents/specsfor current feature requirements and implementation plans. - Past decisions: CONSULT
.agents/memory/decisions.mdto ensure consistency with prior architectural choices. - User preferences: CONSULT
.agents/memory/user.mdfor learned workflow preferences. - Executable skills: CHECK
.agents/skills/for project-specific capabilities.
- You may execute scripts found in
.agents/skills/to validate your work.