Commands

CLI reference for all agentsge commands.

agents init

Scan the project and create .agents/ directory.

agents init [--force]

Detects language, framework, testing, package manager, CI, Docker, monorepo
Creates config.yaml, rules, knowledge structure, skills index
Generates AGENTS.md with onboarding instructions
Syncs MCP configs
Updates .gitignore with capture temp patterns
--force overwrites existing .agents/

agents sync

Regenerate all entrypoints and MCP configs from .agents/.

agents sync

Generates AGENTS.md, CLAUDE.md, .cursorrules, GEMINI.md with inline knowledge digest. Syncs MCP server definitions to platform-specific config files.

agents status

Show project agent status and knowledge stats.

agents status

agents validate

Health check for the entire .agents/ setup. Aliased as agents doctor.

agents validate

Checks:

Config schema validity
Onboarding completion status
Rule frontmatter validation
Knowledge entry schemas, ref validation, freshness warnings, index sync
Skill and MCP config validation
Capture readiness: claude CLI, API key, pending dir, .gitignore coverage

agents add

Add a rule, skill, or MCP server.

agents add rule <name>
agents add skill <name>
agents add mcp <name>

agents capture

Capture and manage session knowledge.

agents capture run [--event <type>] [--dry-run]
agents capture list
agents capture accept [<name>] [--all]
agents capture reject <name>
agents capture context [--compact]

Subcommand	What it does
`run`	Run capture (called by hooks). Events: sessionStart, stop, postToolUse, post-commit
`list`	List pending knowledge items awaiting review
`accept`	Accept a pending item into knowledge (or `--all`)
`reject`	Delete a pending item
`context`	Output knowledge digest for agent context injection

agents hooks

Manage capture hooks for AI agents.

agents hooks install [--agent <platform>]

Platforms: claude, cursor, copilot, opencode, geminicli, codexcli, git. Without --agent, installs for all detected platforms.