The ChunkHound docs are a bit confusing for making it available as an MCP server for Codex. How exactly do you do it? I got up to the indexing step and now need to let Codex be able to use it.