Add a second agent with its own workspace and model
agents & sessions
// what it does
openclaw agents add <id> creates an isolated agent; --workspace sets its files, --model its default, and --bind wires it to a channel in one step. Each agent's workspace defaults to ~/.openclaw/workspace-<id> and its state to ~/.openclaw/agents/<id>/agent, with SOUL.md, AGENTS.md, and USER.md driving its persona.
// shell
$ openclaw agents add work --model anthropic/claude-sonnet-4-6 --bind telegram$ openclaw agents delete work// gotcha
Never reuse agentDir across agents — shared state collides auth profiles and the SQLite session store. Each agent must have its own directory or logins bleed between them.