What problem does it solve? Managing multiple command-line AI coding agents means juggling separate configs, API keys, and context files. This Skill centralizes profile switching between Codex, Claude Code, OpenCode, Gemini CLI, local models, and OpenAI-compatible gateways while keeping one shared memory layer across all agents. ## Core Features & Use Cases - Profile & Recipe Management: Switch active CLI AI profiles with ai-use, install one-step recipes for stacks like OpenCode + OpenRouter, Claude Code, DeepSeek, Ollama, and LM Studio, and manage API presets for 20+ providers. - Shared Cross-Agent Memory: Store global, project, and session context in neutral files exposed through AI_CLI_MEMORY, with tagging, dedupe, compaction, and secret auditing. - Multi-Agent Workflows: Run several agents in one terminal workspace via tmux or Windows Terminal, install agent bridge files so /switch claude works inside a running agent, and pass handoff notes between agents. - Use Case: A developer working in a project runs ai-lite to install agent bridges, then switches between Codex for implementation and Claude for review using ai-workspace switch claude, with both agents reading the same shared memory context. ## Quick Start Ask the AI to set up the CLI model switcher by running the installer and then switching the active profile to a recipe like opencode-openrouter.