lazy system agent
This page is generated from lazy system agent --help.
Usage: lazy system agent [status|set <agent>|set-key <agent>|clear-key <agent>]
See which agent PROFILES are ready and switch the project's default.
An agent is a profile — a harness, a model, an upstream and a credential. The
built-ins (claude-code, codex, cursor, pi) are joined by any [agents.<name>]
block in lazy.toml, and every <agent> below is one of those names.
Subcommands:
status (default) Readiness per profile: harness binary + version, model,
upstream (probed for reachability), credential, and
container capability. Marks the current default.
set <agent> Switch the project default ([agent] agent_id in
lazy.toml, comments preserved). Takes effect on the
next task launch — no daemon restart.
set-key <agent> Store the API key the profile's credential slot bills,
in the same store `lazy auth set` writes (OS keychain
where available, else a 0600 file in the per-project
daemon dir). Kept OUT of the project root: task
containers mount the repo read-only, so an in-repo key
is readable by every agent on the project. Reads the key
from an interactive masked prompt or piped stdin — never
from an argument, which would land in shell history.
Also picked up on the next launch. Human/CLI-only — no
MCP counterpart.
clear-key <agent> Remove that stored key.
Examples:
lazy system agent # readiness view
lazy system agent set cursor # tasks default to the Cursor CLI
lazy system agent set cheap-codex # ... or to a profile this project configures
lazy system agent set-key cursor # prompt for the key (masked)
lazy system agent set-key codex # store an OpenAI API key for Codex tasks
pass show cursor | lazy system agent set-key cursor # piped stdin