Minimal AI coding agent for your terminal. Agentic loop, 13 local tools, streaming Ink TUI, 7 providers behind one UI.

This index is the entry point. The README stays focused on evaluate, install, and first run. Everything deeper lives here.

Start here

• Getting Started - install, key setup, first run, quickstart path
• CLI and TUI - slash commands, keyboard, status line, autocomplete
• Core concepts - map of the ideas, with links to the guides

Core concepts

• Tools - the 13 local executors, caps, approval classes, background tasks
• Agent loop - how one turn runs, what bounds it, how to steer it
• Providers and Models - 7 providers, endpoints, key resolution, model pickers
• Permissions and Modes - normal/yolo, trust, allow/deny rules
• Skills - discovery, frontmatter contract, precedence, auto-invoke

Sessions and context

• Sessions - persistence file, resume, clear, rewind
• Task lists - session checklists for multi-step work
• Compaction and Token Display - auto-compact threshold, manual compact, footer format
• Configuration - env vars, auth file, AGENTS.md layering

Build and fix

• Development - scripts, project structure, tests, build output
• Troubleshooting - common failures and what to check first

Internals

• Architecture - module map of the agent for contributors

Roadmap

Not available yet. Each page names the existing alternative:

• Subagents - status and what to use instead
• MCP - status and what to use instead
• Extensions - status and what to use instead

Agent docs (existing)

Project conventions the agent itself loads at runtime (kept in the atom agent repo, not published here):

• AGENTS.md - agent instructions loaded into the system prompt
• agents/issue-tracker.md - local markdown issues under .scratch/
• agents/triage-labels.md - canonical triage roles
• agents/domain.md - CONTEXT.md plus ADR conventions