Reference

Find a specific guide or setting. For the shortest path to a running round, see Getting started.

Install: the short installation guide
Installation reference: plugins, release binaries, verification, updates and removal
Getting started: set up your project and run a parallel round
How it works: orchestrator, workers and merge checks
Where coding agents go wrong: seven common mistakes, enforced checks, written rules and remaining gaps

All walkthroughs: complete examples for new projects, existing projects and rounds.
Greenfield: what a project holds after a brief. The files /rota-vision, /rota-plan, /rota-capture and /rota-learn leave behind in a new project, with the plan, milestones and knowledge entries shown in full.
Brownfield: what rota adds to an existing project. The project map, config choices, backlog, decision and knowledge entry that appear when rota joins a codebase with history.

Worker setup: hosts, accounts, tracker access and checks
A round cycle: shape the slate, run a round, wind down, release
Parallel rounds: an orchestrator, standing workers in worktrees, the merge gate, solo mode
Unattended rounds: hooks, statusline, keepalive, usage limits, the orchestrator switch
Doctor and reap: check the machine before a round, clear leftovers after
Codex workers: run Codex as a worker in a round
OpenCode workers: run OpenCode, and any model it can reach, as a worker in a round
Orchestrator harnesses: run the orchestrator in Claude Code, Codex, Hermes or opencode
Skills in Codex: install and call the skills from Codex

Capturing work: /rota-capture, mixed input, related links, detail files
Picking work: /rota-work (no argument), /rota-work --preview
Removing work: /rota-capture --remove, dry-run preview, batch removal, safety semantics

Running work: /rota-work parallel cycles, branch vs worktree isolation, capture then work
Debugging: /rota-debug systematic cycle
Pausing and resuming: /rota-pause, recovering after /clear
Parallel work: worktree mode, concurrent /rota-work sessions
Support machine: a second machine in the same round, pinning, claim labels, hand-back

Review and ship: /rota-review single-pass review and /rota-ship gates (second-opinion, QA)
Product QA: /rota-qa per-target strategy files and the ship.qa gate
Rolling back a cycle: /rota-ship --undo guided rollback, dry-run preview, manual confirmation
Learning: /rota-learn and KNOWLEDGE.md, including --term <name> for the project Glossary
External knowledge: a shared git repo of platform and technology learnings, read and written by every project
Decisions: /rota-decide and hard-boundary commitments in DECISIONS.md

Vision and plans: /rota-vision, /rota-plan, milestones
Brainstorming a design: per-item design exploration with /rota-brainstorm, before /rota-plan
Spikes: throwaway feasibility experiments via /rota-spike

Configuration: every key in .rota/config.json and what it does
Autonomy levels: how off, roundtrip and auto change skill chaining
Issue backend: backlog on GitHub/GitLab issues, setup, labels, milestones, rota migrate issues
Umbrella mode: coordinator at umbrella, work in sub-repos (M02 V1)

Safeguards and their limits: the complete enforcement details and remaining gaps
Architecture: how the CLI, the skills and .rota/ fit together
The .rota/ folder: files and directories created by rota init
rota verb reference: every rota verb, with conventions and exit codes
Configuration options: every config key and option label, set via rota config set
Tracker labels: label definitions, project declarations and dry-run sync
Upstream issues reference: GitHub/GitLab issue backend and round-trip closing
Project check: what rota init check verifies, plus exit-code meanings

Rounds on rota itself: the gate, repo rules and roster for contributors
Skill-rule audit, October 2026: the complete never/must inventory, enforcement gaps and observed breakage
Release signing: generating or rotating the minisign key, the Actions secret, verifying by hand
Market review, October 2026: what comparable orchestrators’ users ask for, where rota stands, and the ranked gaps behind issues #574 to #587; refreshed 2026-10-27 with gstack (#2170 to #2174) and new entrants
Plain orchestrator comparison, October 2026: what a bare Claude Code session would do to run a team against what rounds do, and the six gaps behind issues #897 to #902. The re-judged user version is the current one

How rota compares: one page per tool, with the version checked
rota vs GSD: spec-driven phases from discuss to ship
rota vs Superpowers: a skills methodology for many harnesses
rota vs OpenSpec: living specs and delta change folders
rota vs ralph: a fresh-context loop over a prd.json
rota vs Conductor: a Mac app for agents in workspaces
rota vs Orca: a desktop app for fleets of agents
rota vs Vibe Kanban: a kanban board that runs agents
rota vs Claude Squad: a TUI for agents in tmux sessions
rota vs Gas Town: a Mayor, polecats and a merge queue
rota vs Octo: several AI providers on one question
rota or a bare session: a row-by-row comparison with a plain Claude Code session that runs its own subagents