A lightweight AI agent research platform. Small enough to understand (~6k lines), modify, and experiment with.

Getting Started¶

• Getting Started — install, initialize, configure, first chat

Core Concepts¶

• Agent Loop — ReAct loop, sessions, streaming, token budget, error recovery
• Tools — filesystem, shell, web, browser, KB, cron, spawn
• Knowledge Base — persistent memory with topics, search, sync
• Skills — behavioral instructions, self-improvement pipeline, a11y, test automation
• Sub-agents — background task execution, depth tracking, context anchoring
• Sub-agents Guide — practical guide to writing good sub-agent tasks
• Workspace — identity, personality, behavioral templates, TODO tracking

Operations¶

• Scheduling — cron jobs, heartbeat, daemon mode
• Dashboards — TUI terminal dashboard, web dashboard
• JSON Channel — programmatic stdin/stdout protocol
• Configuration — full config reference

Reference¶

• Architecture — module map and system design
• Roadmap — completed phases and future plans