Kimi Code is Moonshot AI’s terminal coding agent: it reads and edits code, runs shell commands, searches the web, and plans autonomously. The CLI is open source under MIT, and it replaced Moonshot’s original Kimi CLI, which is now archived.
models and pricing
Kimi Code runs Moonshot’s own models: K3.1 Pro is the flagship, alongside K3, K2.8 Preview, and K2.7 Code Highspeed, with up to 1M context on higher tiers. It is bundled into Kimi memberships from about $19/mo (the Plus plan) under a rolling 5-hour rate window and a monthly quota, and a Moonshot AI Open Platform API key also works for sign-in. The docs say other providers can be configured: Anthropic, Google, and OpenAI-compatible services.
what it’s good at
Kimi Code treats the shell as a first-class surface: a shell mode,
entered with !, runs terminal commands inside the conversation and
keeps their output in context. It ships built-in subagents
for coding, exploration, and planning, lifecycle hooks, MCP support with
a conversational config flow, editor integration over ACP for Zed and
JetBrains, and video input.
Kimi Code in crystl
Kimi Code runs in a crystl shard with its own model selection, activity visibility, and real approval cards for its tool calls. Crystl treats its compatibility as partial because its hook surface can only deny, and because transcript reading and resume by id are both still unwired.
The profile describes what crystl has tested; it never supplies a command, hook script, or launch setting. Those remain local to the Crystl installation.