apm targets

Show which agent harnesses APM will deploy to from the current project, and why.

Synopsis

apm targets [OPTIONS]

Description

apm targets prints the canonical target table for the current directory: every supported harness, whether APM considers it active, the filesystem signal that activated it, and where compiled output will land.

Resolution order matches apm compile:

1. --target / --all on the command line (not applicable to apm targets itself, but reflected by compile and install).
2. targets: field in apm.yml.
3. Auto-detection via filesystem signals (see Detection signals).

Use this command before apm compile or apm install to confirm what auto-detection resolves to. If APM lists a target you do not intend (for example, CLAUDE.md is project documentation, not a Claude Code config), pin targets: explicitly in apm.yml.

Subcommands

apm targets is a Click command group, but no subcommands ship today. Invoking apm targets without arguments prints the resolved-target table.

Planned

Subcommands such as apm targets add are reserved on the group for future use. They are not implemented yet.

Options

Flag	Description
--json	Emit machine-readable JSON instead of the table. One object per canonical target with target, status, source, deploy_dir, needs.
--all	In --json mode, include the agent-skills meta-target row (excluded by default). No effect on table output.

The agent-skills meta-target is a multi-harness fan-out for shared .agents/skills/ output. It is not a harness and is excluded from the default table.

Examples

Show the resolved target table:

apm targets

Sample output in a project with CLAUDE.md and .cursor/:

  TARGET       STATUS     SOURCE                                   DEPLOY DIR
  ------------ ---------- ---------------------------------------- ----------
  claude       active     CLAUDE.md                                .claude/
  copilot      inactive   needs .github/copilot-instructions.md    .github/
  cursor       active     .cursor/                                 .cursor/
  codex        inactive   needs .codex/                            .codex/
  gemini       inactive   needs GEMINI.md                          .gemini/
  opencode     inactive   needs .opencode/                         .opencode/
  windsurf     inactive   needs .windsurf/                         .windsurf/

Machine-readable form:

apm targets --json
apm targets --json --all

Detection signals

Auto-detection walks the project root for these markers. The first match per target is enough to activate it.

Target	Signal(s) APM looks for	Deploy directory
claude	.claude/ directory, or CLAUDE.md file	.claude/
copilot	.github/copilot-instructions.md file	.github/
cursor	.cursor/ directory, or .cursorrules file (legacy)	.cursor/
codex	.codex/ directory	.codex/
gemini	.gemini/ directory, or GEMINI.md file	.gemini/
opencode	.opencode/ directory	.opencode/
windsurf	.windsurf/ directory	.windsurf/
agent-skills	Meta-target; never auto-detected. Opt in via targets: in apm.yml or --target agent-skills on apm compile.	.agents/skills/

Notes:

• Detection is filesystem-only. APM does not inspect file contents to decide whether a marker is “real”.
• A CLAUDE.md written as documentation will still activate the claude target. Pin targets: in apm.yml to override.
• If no signals are found and apm.yml declares no targets:, apm targets prints the full table with every row inactive and an info hint to create a harness config or declare targets: explicitly.

Exit codes

Code	Meaning
0	Table or JSON printed (including the all-inactive case).

Related

• apm install — uses the same target resolution to decide which harness configs to wire.
• apm compile — compiles primitives for the targets shown here.
• Targets matrix — per-target output layout and feature support.
• Concepts: primitives and targets