Documentation

Atlas Tasker Docs

Everything you need to run Atlas Tasker. Tickets are plain Markdown, every change is an audited event, and one tracker binary serves the CLI, the TUI, Atlas Home, and MCP. Ordinary onboarding is tracker init then tracker. v1.16 adds native client skills and everyday browser management. Remote backup is optional and never guesses your Git origin.

Start here

New to Atlas? Two pages get you from nothing to a working board with an agent on it.

How Atlas thinks

Three ideas explain most of the design. Once they click, every command makes sense.

  • The repo is the database. Tickets are Markdown files under projects/, and every change appends an event under .tracker/. A small disposable index keeps queries instant, rebuilds itself from the files when it is missing or behind, and tracker doctor --repair rebuilds it on demand. The FAQ covers the storage internals honestly.
  • Every write is attributed. CLI mutations resolve --actor, TRACKER_ACTOR, then actor.default, and fail before writing if identity is missing. Reasons explain why; tracked MCP writes and protected operations require them.
  • Machines are first-class users. Commands support --json, exit codes are stable, and an MCP server exposes typed tools behind escalating profiles. Interactive prompts are confined to workspace and integration setup; JSON and non-TTY paths stay non-interactive. Humans get the same tracker with nicer pixels.

The manual

Deeper reference

These pages distill the essentials. The exhaustive, always-current reference ships with the source:

  • Command reference: every command and flag, including governance, sync, signing, backups, and goal manifests.
  • AGENTS.md: the operating guide written for agents themselves. Point your agent at it, or install it with tracker integrations install.
  • The full docs tree: architecture, threat model, storage transactions, and release engineering.