FAQ

What does atl stand for?

AgentTeamLand. The CLI, the org, and the ecosystem share the name.

Does atl replace Claude Code?

No. atl is a delivery layer for the files Claude Code already reads from .claude/. You still run Claude Code exactly as before; atl just makes it easier to get a solid configuration in place — and keeps it improving on its own.

How is this different from just copying files between projects?

Three ways:

1. Versioning. Teams tag SemVer releases. atl update pulls the latest published version of each installed team, so improvements fan out to every install.
2. Scope. A team installs at a scope — global (~/.claude/) or project (<project>/.claude/). When both carry an asset with the same name, the project copy shadows global. Hand-copied files have no such axis.
3. A self-running learning loop. As your agents work, they accumulate gains — new learnings, sharpened skills. atl captures them automatically into a durable queue, /drain folds them into agent knowledge bases, and atl promote / atl publish circulate them outward. Copying files never gets better on its own.

Do I need to run atl to use Claude Code?

No. Claude Code works fine without atl. Use atl when you want a reproducible, shareable setup — solo projects with a hand-rolled .claude/ are still perfectly valid.

Can I install more than one team in the same project?

Yes. Each install adds its own copies into .claude/. If two teams ship an agent with the same name, the most-recently-installed team's version wins (it overwrites the earlier copy) and atl prints a one-line warning. This is collision handling, not inheritance — each team is installed independently. Use atl list to see what's installed at each scope.

Where do teams come from? Can I install from a private repo or a Git URL?

atl install is catalog-only. It takes a <handle>/<team> reference, resolves it against the GitHub-backed catalog (built from public repos tagged with the atl-team topic), fetches the source as an ephemeral HTTPS tarball, and copies the team's agents/, skills/, and rules/ into the scope's .claude/.

There is no install from a private repo, an arbitrary Git URL, SSH, or a local path — those were v1. If you want a team installable, make its repo public and tag it atl-team (or run atl publish from the repo). See atl search for how the catalog is queried.

What if my atl version is too old for a team?

A team's team.json can declare a requires.atl minimum to signal the version it expects. To stay current, run atl update (it also refreshes the atl binary's bundled core), or re-run the install script:

curl -fsSL https://raw.githubusercontent.com/agentteamland/atl/main/scripts/install.sh | sh

What happens to atl's on-disk state if I delete a project?

Nothing global is affected. atl keeps two kinds of state, and they're separate:

• Team assets live in Claude Code's own .claude/ directories. Deleting a project deletes that project's .claude/; global assets in ~/.claude/ are untouched.
• atl's bookkeeping lives under ~/.atl/ (global) and <project>/.atl/ (project) — the cached catalog (index.json), the learning queue (queue.db), pins, and per-team install manifests. The global ~/.atl/ survives; the project's <project>/.atl/ goes with the project.

There is no shared clone cache to clean up — sources are fetched as throwaway tarballs at install time, never kept on disk.

Can I install a team without running atl (by hand)?

Yes — atl just automates a copy. Fetch the team repo, then copy its agents/, skills/, and rules/ into the target .claude/ yourself. There's no inheritance or excludes resolution to replicate, and no persistent cache to populate. The only thing you'd lose is the install manifest atl writes (see below), which atl update and atl doctor rely on for refresh and self-heal.

Where does atl keep the list of installed teams?

In a per-team manifest, one JSON file per team per scope, at <scope>/.atl/installed/<handle>__<name>.json (<scope> is ~/.atl for a global install, <project>/.atl for a project install). Each manifest records the team's handle, name, version, scope, source, install time, and a files map of every written path to its SHA-256 baseline. atl update's auto-refresh and atl doctor's integrity check read those baselines. Editing manifests by hand is unsupported — use atl install / atl remove.

Does atl send any telemetry?

No. atl is a local tool: it fetches teams over HTTPS, reads the catalog, and writes copies into .claude/. There's no phone-home.

Is this an Anthropic product?

No. AgentTeamLand is an independent open-source project that works with Anthropic's Claude Code. MIT-licensed. No commercial affiliation.

How do I contribute?

• Publish a team. Tag the team's repo with the atl-team topic, or run atl publish from the repo. The catalog picks it up automatically — there's no registry and no submission PR.
• Improve the CLI, the core rules/skills, the first-party teams, or these docs. They all live in the agentteamland/atl monorepo. PRs welcome; every docs page has an "Edit this page on GitHub" link.
• Open issues. Bug reports and feature requests go on agentteamland/atl.

My question isn't here.

Open an issue on agentteamland/atl with the faq label. If it's a common question, it'll end up on this page.