Kimi Code is Moonshot AI’s agentic coding tool — a CLI and IDE-side agent in the same product family as Claude Code or Codex — and its pricing works like a bundle, not a product. There is no “Kimi Code plan”: you buy a Kimi membership, and each paid tier includes a Kimi Code credit multiplier alongside the chat product, agent credits, and Moonshot’s research-preview features. The musical tier names (Moderato through Vivace) map to $19, $39, $99, and $199 per month, with annual billing cutting the effective monthly price by roughly 20%.
Two things stand out against the US-tool pricing in the rest of this tracker. First, the entry price is aggressive: $19/mo buys access to a frontier-class model (Kimi K3 as of July 16, 2026) in an agentic coding harness, undercutting the typical $20 entry tiers while claiming benchmark parity with much more expensive models. Second, the limits are unusually opaque. Moonshot publishes credit multipliers (1×, 5×, 15×, 30×) rather than token budgets or request counts, and early reports suggest heavy agentic tasks can drain a low tier quickly — K3 is a token-hungry model, and the credits meter that consumption. Our advice matches what we said in the K3 review: judge cost per completed task on your own workload, not the sticker price.
The escape hatch from the subscription is the API: $3 per million input tokens, $15 per million output, $0.30 cache-hit, flat at any context length. That’s metered through the Moonshot platform (or resellers like OpenRouter) and billed separately from memberships — a membership does not include API usage. The free Adagio tier, for what it’s worth, is a chat-product trial only; it carries no Kimi Code access at all.