Integration · Codex CLI (OpenAI)

Codex CLI ⇄ SwiftyClip

Codex speaks MCP. Once SwiftyClip is registered in ~/.codex/config.toml, any Codex prompt can reach the ten clip.* tools from the terminal.

1. Install SwiftyClip

Grab the notarized DMG from /download. Launch once — that boots the local bridge that Codex will reach over stdio.

2. Register the MCP server

Edit ~/.codex/config.toml:

[mcp.servers.swiftyclip]
command = "swiftyclip"
args = ["mcp"]

3. Prompt Codex

$ codex exec --full-auto "Use swiftyclip to process ~/Movies/ep.mp4. Score segments, render the top 3 with bold captions to ~/Desktop/."

Codex chains clip.ingest → clip.analyze → clip.scoreSegments → clip.render autonomously. Use --full-auto to skip per-tool confirmation prompts during long batches.

Tip: CI workflows

Because Codex is scriptable, you can wire SwiftyClip into a local build loop: a watch process that triggers Codex on new files in ~/Recordings/, and SwiftyClip clips them overnight. Pair with the overnight-schedule for a lights-out workflow.