One binary, no Node.js: a terminal agent that watches video

What it does

Kimi Code CLI is MoonshotAI’s terminal-based coding agent. It reads and edits files, runs shell commands, searches your project, fetches web pages, and decides what to do next based on the results. It works out of the box with Moonshot’s own Kimi models, but you can point it at other compatible providers if you prefer.

The interesting bit

The TUI is built for long, focused sessions rather than quick one-off questions, and it will accept actual video files—drop in a screen recording and the agent watches it. That’s a rare input modality for a terminal tool, and it sidesteps the usual ritual of writing paragraphs to describe a visual bug.

Key highlights

• Ships as a single binary with no local Node.js dependency for end users.
• Boots in milliseconds, which matters more than it should for a tool you’ll restart often.
• Includes built-in subagents (coder, explore, plan) that work in isolated contexts so the main conversation stays tidy.
• Configures MCP servers conversationally via /mcp-config instead of hand-editing JSON.
• Exposes lifecycle hooks so you can gate risky tool calls or trigger your own automation at key decision points.

Caveats

• The README mentions “other compatible providers” but never lists which APIs or standards are supported, so compatibility is unclear.
• Video input is mentioned as a feature, but there is no detail on formats, limits, or how it is processed.

Verdict

Worth a look if you want a long-running terminal agent with built-in task delegation and video support. Skip it if you need a clear list of supported non-Moonshot providers before installing.