klim agent launch

klim agent launch opens an interactive agent session via the provider’s own CLI, but you don’t have to remember each provider’s flag syntax. Pick the entity (skill / plugin / saved session), and klim builds the right command line and either runs it or prints it for review.

Usage

klim agent launch [flags]

Flags

Flag	Description
--provider	Required when ambiguous: claude-code, copilot-cli.
--skill <name>	Launch with a specific skill loaded.
--plugin <name>	Launch with a specific plugin enabled.
--session <id>	Resume a saved session (use the full id from klim agent session list).
--print-only	Print the launch command instead of executing it — useful for scripts or remote shells.
--cwd <path>	Override the working directory for the launched session.

Examples

# Start a Claude Code session with the summarize skill loaded
klim agent launch --provider claude-code --skill summarize

# Resume a saved session by id (URL-escaped path)
klim agent launch --session "claude:home%2Fuser%2Frepo"

# Just show me the command — don't execute it
klim agent launch --provider copilot-cli --plugin test-runner --print-only

How it works

klim doesn’t shim the provider — it locates the provider’s binary on PATH, builds the right argv (passing skill / plugin / session refs in the provider’s native format), and execs into it. The session inherits your TTY, environment variables, and signal handlers exactly as if you’d typed the provider command yourself.

--print-only is the escape hatch when you want to embed the launch command in a shell script, a tmux config, or an IDE task without actually starting an interactive session right now.