API guide
DeepSeek API setup for a terminal agent
The DeepSeek API is the engine. The terminal agent is the operating layer around it. A good setup keeps keys out of chat, routes models intentionally, and makes cost visible while work is happening.
For users who are ready to connect an API key and want fewer setup surprises.
Local setup checklist
DeepSeek-TUI can prompt for an API key on first launch, store it in local config, or read it from environment variables. For team use, document which path is approved and how key rotation works.
A hosted workspace can reduce setup variance by keeping the purchase path, workspace defaults, support flow, and model route recommendation in one place.
- Create the key from the official DeepSeek platform.
- Avoid pasting secrets into prompts or issue threads.
- Run the built-in doctor command after setup.
- Use provider-specific config only when you know why you need it.
Cost controls to care about
The API bill is easier to manage when the agent reports the actual model used, token usage, reasoning output, and cache behavior. DeepSeek-TUI is valuable because cost is visible at the turn level instead of arriving as a surprise later.
Questions worth answering before checkout
Can I use environment variables instead of saved config?
Yes. Environment variables are useful in scripted or ephemeral environments, while saved local config is often smoother for daily terminal work.
Does the hosted plan include the DeepSeek API itself?
The hosted plan covers the workspace layer. Model/API usage can depend on the provider and account terms shown during onboarding.