How do you run DeepSeek in Claude Code?

Claude Code sends every request in Anthropic's Messages format. It reaches another server when ANTHROPIC_BASE_URL is set to that server's address.

DeepSeek serves its models in that same format, at https://api.deepseek.com/anthropic. Ollama Cloud and OpenRouter answer in it too.

So unlike GPT, DeepSeek needs no translation in the middle. What it needs is the right model names, the right window and a key. How GPT differs is in running GPT in Claude Code.

DeepSeek's own guide sets nine environment variables before you start Claude Code:

• the base URL above, and your DeepSeek API key as ANTHROPIC_AUTH_TOKEN
• deepseek-flash[1m] as the main model, and as the Opus and Sonnet model
• deepseek-flash as the Haiku model and the subagent model
• CLAUDE_CODE_AUTO_COMPACT_WINDOW at 786432, and the effort level at max

DeepSeek's endpoint maps Claude names too. A name starting with claude-opus runs on DeepSeek V4 Pro, at V4 Pro's price. Any other Claude name runs on DeepSeek Flash.

Its Anthropic format has gaps. It ignores cache_control and mcp_servers, and it does not take documents such as PDFs.

That works for one person. On a team, it is one DeepSeek key and one set of variables per laptop, and no shared view of what DeepSeek costs. Why a key on every laptop is the expensive part is in keeping API keys off laptops.

An owner connects an Ollama Cloud plan or an OpenRouter key in Selan. DeepSeek V4.1 Flash and V4 Pro are models Selan lists, so an owner also switches each one on at that provider under Settings → Models.

Developers install the selan CLI, sign in as themselves and name the model:

The part before the first colon is the provider. Everything after it is that provider's own model id, passed through as it is.

With --model, Selan tells Claude Code the model's real context window. Selan lists DeepSeek V4.1 Flash at 1,048,576 tokens and V4 Pro at 1,000,000, so a session compacts there and not at 200K. If an owner set a lower context limit for your company, it compacts at that instead.

Both start at high effort, and Selan offers low, high and max for them. V4.1 Flash reads images. V4 Pro takes text only.

To skip the flag, an owner adds DeepSeek to Claude Code's picker under Settings → Model pickers. It then appears in /model. The details are in Run any model.

If that account is out of quota, the request fails rather than moving to the other one.

Selan does not connect DeepSeek's own API. It reaches DeepSeek through Ollama Cloud and OpenRouter. All of this is on Providers and models.

• One usage record. DeepSeek turns land beside Claude turns, per person and per repository. Usage per developer shows what an owner sees.
• One spend limit. An OpenRouter turn is metered, so it counts against a member's limit. A turn on an Ollama flat plan does not. See spend limits.
• No DeepSeek or OpenRouter key on a laptop. Keys stay on the gateway. A developer holds a Selan session you can revoke.
• Codex on the same accounts. selan codex --model ollama:deepseek-v4.1-flash uses the same route. Without the provider in front, Codex means the model itself, and Selan falls through your accounts for it. See Codex.

• Effort and spend. DeepSeek thinks before it answers. On OpenRouter, the thinking is metered too.
• The model must be switched on. A model your company cannot use is refused before Claude Code starts, not at the first request.
• Pick the model by the job. V4 Pro has no image input, so a screenshot in a session needs V4.1 Flash.

Free for up to three seats, and the Free plan needs no card.