All models
Moonshot

Kimi K2.7 Code API: pricing, base URL and quick setup

Kimi K2.7 Code is available on A2Agent as kimi-k2.7-code at $0.950 per 1M input tokens, with a 256K context window, through one OpenAI-compatible endpoint. Coding agents on Moonshot at a 256K window.

Specifications and pricing

Model idkimi-k2.7-code
PlatformMoonshot
Context window256K
Tagged forchat, coding
Discount
Input / 1M tokens$0.950
Output / 1M tokens$4.00

Prices synced with platform billing (USD per 1M tokens).

Quick setup

curl https://a2agent.me/v1/chat/completions \
  -H "content-type: application/json" \
  -H "authorization: Bearer sk-your-a2agent-key" \
  -d '{
    "model": "kimi-k2.7-code",
    "messages": [{"role": "user", "content": "Reply with OK"}]
  }'

Using it from a coding client

Set it as the edit model in Cline, Roo Code or Kilo Code and keep something cheaper for the read-only steps. Most clients allow a model per role, and that split is where the savings live.

How it compares within its family

K2.7 Code is the coding-tagged member of the Moonshot family. It lists at $0.95 per million input tokens and $4.00 output, identical to K2.6, and keeps the same 256K window. The tag is the whole difference the catalogue records. Against K2.5 it costs about half again as much on input for the same window; against K3 it is roughly a third of the price with a quarter of the context. For a coding agent the window usually decides: 256K holds a package and its tests, but a repository-wide pass over a large codebase needs chunking here where a 1M model would not. Pin the id exactly, since kimi-k2.7-code and kimi-k2.6 are different models at the same price.

Other models on this platform

Questions

What makes it a coding model?

It is the Moonshot model carrying the coding tag in this catalogue. K2.5 and K2.6 are tagged for chat and vision, K3 for chat and vision as well.

What is the context window?

256K tokens, the same as K2.5 and K2.6.

Does it cost more than K2.6?

No. Both list at $0.95 per million input and $4.00 output.

Will it handle a whole repository?

A package and its tests, comfortably. For a repository-wide prompt, a 1M model such as K3 or GLM-5.3 avoids the chunking.

Trial credit on sign-up, billed per token, no minimum spend.