deepseek support

[iron vault]

already existing prs, very easy to do, but unmerged, and neccesary imo

[topaz tree]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[iron vault]

so krill, either you make a pr or you help me get one of the prs pushed into mainline

it adds support for deepseek in openclae

*openclaw

@jagged glacier

Hey @jagged glacier pls respond

[jagged glacier]

Hey! I’m here — what do you need help with?

If it’s an OpenClaw/DeepSeek setup issue, paste:

• what you’re trying to do
• the exact error text
• your openclaw --version
• (if relevant) the config snippet you’re using (redact keys)

[iron vault]

Deepseek isn't even an option to use

Through the official deepseek api

There are unmerged fully working prs that implement this correctly

I'm not able to use deepseek by manually settinh it though

[jagged glacier]

Hamb_y — in OpenClaw, “DeepSeek” usually won’t show up as a ready-made provider option unless you’re using a provider that already exposes DeepSeek models (or you add it as a custom provider).

Most common reasons it “isn’t even an option”:

1. You haven’t configured any provider that offers DeepSeek models
   • Easiest path is often OpenRouter (it includes DeepSeek models), or Venice, or local Ollama.
2. You have an allowlist enabled (agents.defaults.models) and DeepSeek isn’t in it
   → then /model list won’t offer it and selecting it will fail with “Model is not allowed”.

Can you paste:

• openclaw --version
• output of openclaw models status (or a screenshot)
• and tell me where you’re expecting to see it (CLI openclaw models list, in-chat /model, or dashboard)?

If you’re trying to use DeepSeek’s own API key directly, you’ll need to add it as a custom provider under models.providers (OpenAI-compatible baseUrl + apiKey) — docs section: https://docs.openclaw.ai/gateway/configuration (search for “models (custom providers + base URLs)”).

[jagged glacier]

DeepSeek won’t show up as a built-in “option” unless you’ve configured a provider that exposes DeepSeek models (or you add DeepSeek as a custom provider). Also: if you have a model allowlist enabled, /model won’t let you select anything that isn’t listed.

Can you paste the exact error you get when you try to set it (and also openclaw --version)?

In the meantime, these 2 quick checks usually reveal what’s wrong:

openclaw models status
openclaw models list --all | grep -i deepseek

What the results mean

A) grep shows nothing
→ OpenClaw currently has no configured provider that offers DeepSeek models, so you can’t “manually set” it because OpenClaw doesn’t know that model exists yet.

Fix: configure a provider that has DeepSeek models (commonly OpenRouter or Venice) or add a custom provider in models.providers (OpenAI-compatible baseUrl + apiKey + models list). Docs for provider configs/custom providers:

[iron vault]

I tried it. Not on my pc rn so can't paste. I'm talking about the official deepseek api, not openrouter middlemen