Practical Notes
- Setup time ~8 minutes (install + set one API key + first launch)
- Ships built-in slash commands for plan/memory/MCP management (e.g.,
/plan,/memory,/mcp) - GitHub stars/forks (verified) for momentum signal: see Source & Thanks
A productive way to use omp is to treat it as a repeatable harness rather than a “chat UI”:
- Start in your repo root and keep the agent’s working directory stable.
- Use the built-in slash commands to manage state explicitly: switch models, compact context, and toggle MCP servers only when needed.
- Keep your runs reviewable: export sessions, and keep changes small enough to validate quickly.
Two tips that save time:
- If your terminal doesn't support the Kitty keyboard protocol, configure it first (otherwise modifier keys can behave inconsistently).
- Prefer a single provider/API key at the start. Add more providers only after you’ve stabilized your workflow.
FAQ
Q: What is omp?
A: omp is the interactive terminal UI entrypoint for oh-my-pi's coding agent harness.
Q: Do I need Bun? A: The repo recommends Bun for installation in some paths, but the installer script can install a prebuilt binary when available.
Q: How do I connect MCP servers?
A: Use the in-chat /mcp command to add/toggle MCP servers as you need them.