Moonshot
Kimi K2.7 Code API
Kimi K2.7 Code is Moonshot's coding-focused route for agentic programming, repository-level edits, and multi-step developer workflows. It is for developers who want a model to do real engineering work in a controlled loop: inspect code, use tools, make a focused change, and explain how to verify it.
The model is most useful when the agent harness is well designed. Give it a task with acceptance criteria, constrain the filesystem and commands it can access, and let tests—not fluent prose—decide whether an edit is correct. That turns repository-level capability into an auditable workflow.
It can be selected through the same API configuration as the rest of GlideflowAI's menu. This makes it easy to run a code-task suite against Kimi, GLM, Qwen, or Claude candidates without swapping SDKs or credentials.
What Kimi K2.7 Code is good for
- Repository-level feature work with an issue, code context, and test command.
- Agentic bug fixes that inspect failures before editing the smallest relevant area.
- Refactoring a module while maintaining tests and documenting the migration.
- Generating a patch plan after exploring unfamiliar project structure.
- Multi-step developer workflows with tool calls for search, tests, and diffs.
New to the gateway? Start with the quickstart, then review authentication before putting a key in an application.
Before a production launch, keep credentials server-side, record each selected model, token use, and validation result, and define a per-run budget. Those traces let you compare models on the workflow that matters, diagnose failures with evidence, and change a routing rule without guessing when product requirements or traffic change.
When to choose Kimi K2.7 Code
Choose Kimi K2.7 Code when repository edits and agentic programming are the primary job. Kimi K2.6 is a broader long-form reasoning and coding option; GLM-5.2, LongCat-2.0, and Claude coding routes are useful alternatives to evaluate on your actual repositories.
Compare the current token rates on Pricing and test the nearest alternatives on your real prompts. The related models below are a useful starting set for that evaluation.
Pricing
USD per 1M tokens. No hidden markup.
Input
Output
Call Kimi K2.7 Code in seconds
It's OpenAI-compatible. Point your existing client athttps://api.glideflowai.com/v1and set the model tokimi-k2.7-code.
curl https://api.glideflowai.com/v1/chat/completions \
-H "Authorization: Bearer sk-xxx" \
-H "Content-Type: application/json" \
-d '{
"model": "kimi-k2.7-code",
"messages": [{"role": "user", "content": "Hello"}]
}'FAQ
Can Kimi K2.7 Code work in a coding-agent harness?
It is positioned for agentic programming and repository edits, so it is a good candidate to test in compatible harnesses. Start in a disposable or branch-based environment and verify the tool configuration before granting write access.
How do I make repository changes safer?
Require a plan, limit the changed paths, run tests, inspect the diff, and keep commits or pull requests as the review boundary. A model should not bypass your normal engineering controls.
What model string should my code send?
Use kimi-k2.7-code as the exact model ID. The display name and the API identifier differ only where the catalog needs a stable machine-readable name.
How should I price an agentic coding task?
Count the full loop: repository context and tool results are input, while plans, patches, and explanations are output. Set per-task budgets because repeated search and test iterations can dominate a simple one-shot estimate.