Cline (and its fork Roo Code) is the most popular open-source coding agent for VSCode: it plans, edits files, runs terminal commands, and iterates — with any model you point it at. That flexibility makes it the easiest way to run agentic coding on Chinese models through NovAI.
OpenAI Compatible.https://aiapi-pro.com/v1kimi-k2.7-code or glm-5.3.Roo Code is a Cline fork with per-mode model routing (Architect / Code / Ask). The provider setup is identical: OpenAI Compatible → base URL https://aiapi-pro.com/v1 → NovAI key. A useful pattern: Architect mode on glm-5.3 (1M context for repo-wide planning), Code mode on kimi-k2.7-code (agent-tuned edits), Ask mode on free glm-4.7-flash.
| Model | Input $/1M | Output $/1M |
|---|---|---|
| kimi-k2.7-code (agent-tuned) | $0.67 | $3.40 |
| glm-5.3 (1M context flagship) | $1.25 | $4.00 |
| deepseek-v4-pro | $0.57 | $1.15 |
| doubao-seed-2.0-code | $0.448 | $2.24 |
| glm-4.7-flash (free forever) | $0 | $0 |
For long agent loops, kimi-k2.7-code and deepseek-v4-pro give the best cost-per-task; for repo-wide refactors that need the full codebase in context, glm-5.3's 1M window is unmatched. Use free glm-4.7-flash for planning and lightweight edits.
Agentic coding in VSCode from $0.57/1M tokens — $2 free credit to start.
In Cline settings choose API Provider 'OpenAI Compatible', set Base URL to https://aiapi-pro.com/v1, paste your NovAI API key, and type the model ID (e.g. kimi-k2.7-code, glm-5.3, deepseek-v4-pro).
Yes. Roo Code is a Cline fork with the same OpenAI-Compatible provider option. You can additionally assign different NovAI models to Architect, Code, and Ask modes.
deepseek-v4-pro at $0.57 per 1M input tokens is the cheapest strong coder on NovAI; glm-4.7-flash is completely free and fine for planning and small edits.