If you’d rather stay in VS Code than live in the terminal, Cline is the way to get an AI coding agent inside the editor; with a visual plan-and-approve workflow that shows you every diff before it lands. It’s free and open source, and it runs on DeepSeek, so the only cost is DeepSeek’s very low pay-per-token rate.
This is the full Cline + DeepSeek setup on Windows, no WSL required since it’s an editor extension. For how Cline compares to terminal agents, see best cheap Claude Code alternatives.
Step 1: Install Cline
In VS Code, open the Extensions panel, search for Cline, and install it. A Cline icon appears in the sidebar.
Step 2: Get a DeepSeek API key
Create a DeepSeek account, add a little credit, and generate an API key.
Step 3: Configure the provider
Open Cline’s settings (the gear in the Cline panel) and set:
Cline settings for DeepSeek
| API Provider | OpenAI Compatible |
|---|---|
| Base URL | https://api.deepseek.com |
| API Key | your DeepSeek key |
| Model ID | deepseek-v4-flash (or deepseek-v4-pro) |
Choose OpenAI Compatible as the provider, paste in the base URL and key, and set the model. Save, and Cline is running on DeepSeek.
Step 4: Plan vs Act
Cline has two modes worth knowing:
- Plan; the model lays out its approach first, so you review before any files change. Great for keeping a cheaper model on the rails.
- Act; the model edits files and runs commands, showing diffs for approval.
Start a task in Plan mode for anything non-trivial, confirm the plan, then let it act.
Pick the model
deepseek-v4-flash: routine coding and faster responses.deepseek-v4-pro: harder reasoning and complex debugging.
Switch the model ID in Cline’s settings when the task changes. DeepSeek discontinued deepseek-chat and deepseek-reasoner on July 24, 2026, so old configurations must be updated.
Pricing
Cline is free; DeepSeek is pay-per-token with no coding plan, plus cache-hit discounts and off-peak windows. Together they make in-editor agentic coding genuinely cheap. Confirm DeepSeek’s current rates on its pricing page; see DeepSeek V4 pricing explained.
Troubleshooting
- Auth errors; wrong key, or extra spaces when pasting. Re-paste carefully.
- Format errors; you used the
/anthropicURL; switch tohttps://api.deepseek.com. - Model not found; the Model ID doesn’t match DeepSeek’s current names.
- No diffs appearing; you’re in Plan mode; switch to Act to apply changes.
DeepSeek + Cline checklist
- Cline extension installed in VS Code
- DeepSeek API key created
- Provider set to OpenAI Compatible
- Base URL https://api.deepseek.com and key entered
- Model set to deepseek-v4-flash; tested in Plan mode
Verify the connection
Ask Cline to inspect a small file in Plan mode before allowing edits. If the request succeeds, switch to Act mode and review the first diff. A model-not-found response usually means the configuration still contains an old DeepSeek alias.
Prefer the terminal? See DeepSeek + OpenCode or DeepSeek + Aider. Want GLM in Cline instead? See GLM + Cline in VS Code.
