Wiring a Claude Code desktop DeepSeek setup just went official: Ollama now ships Claude Desktop support as a third-party gateway, so Anthropic’s harness can run any Ollama model — DeepSeek included, local or cloud — with one toggle back to stock. Here’s the setup and the fine print.
Short answer
- Official (25 Aug): toggle Claude on inside Ollama and the gateway configures itself.
- Any Ollama model behind Claude Desktop — DeepSeek locally, or Ollama cloud for big jobs.
- Ollama: telemetry off by default, strict Zero Data Retention, nothing sent to Anthropic.
- It’s Claude Desktop, not the CLI — and one toggle restores your old setup exactly.
The Claude Code desktop DeepSeek route: what Ollama shipped
On 25 August, Ollama announced official Claude Desktop support: developers can configure Claude Desktop to work with Ollama as a third-party gateway provider. Which quietly means the Claude Code desktop DeepSeek combination people kept asking about now has an official route.
The shape of it: open Ollama, select Claude, toggle it on — Ollama configures the gateway for you. While connected, Claude Desktop can use any model within Ollama, local models running on your own machine or larger ones on Ollama’s cloud. Toggle it off and your previous Claude setup is restored exactly.
DeepSeek’s open models live in Ollama’s library — which is what makes this pairing interesting: Anthropic’s polished desktop harness, DeepSeek’s economics underneath.
Setting up Claude Desktop with DeepSeek via Ollama
- Download Ollama and open it.
- Select Claude in Ollama’s settings and turn the toggle on — the third-party gateway is configured for you. That’s genuinely the setup.
- Pick your models: the settings page lets you choose which model Claude uses for each model option — a DeepSeek model locally if your hardware carries it, or via Ollama’s cloud for the bigger jobs.
- Swap freely. You can keep using Anthropic models too — switching is the toggle, and turning Claude off in Ollama restores your previous setup.
Cloud models for larger coding and research tasks, open models directly on your computer, and your existing Claude setup one toggle away — that’s the whole design.
🔥 Want this set up without the guesswork? Pairing the right brain with the right harness per job is the entire game now — and it keeps getting easier. Inside the AI Profit Boardroom you get the Agent OS as a ready-to-install file, a 30-day roadmap, daily tutorials the same day new tools ship, and four live coaching calls a week where you share your screen and get unstuck. → Get access here
The privacy line, stated by the vendor
Ollama’s FAQ is unusually direct: no prompts or data are sent to Anthropic or Ollama — telemetry is disabled by default, and Ollama states a strict Zero Data Retention policy across its models and services, cloud or local.
Run a DeepSeek model locally through this and you get the combination this session has kept circling: a first-class harness with a brain whose data never leaves the building. That’s a different privacy posture from DeepSeek’s own API, whose terms allow training on what you send — the comparison is in the V4 wiring guide.
What this changes — and what it doesn’t
| Changes | Doesn’t change |
|---|---|
| Claude Desktop can run non-Anthropic brains officially | This is Claude Desktop — not a general Claude Code CLI unlock |
| DeepSeek-class economics inside Anthropic’s harness | Frontier Claude models still lead the hardest reasoning |
| Local-model privacy in a mainstream desktop app | Local hardware still limits which models run well |
| One-toggle switching between ecosystems | Your model-per-job judgement — that’s still on you |
Read the pattern underneath: harnesses and brains are decoupling everywhere — DeepSeek Harness reads Claude’s instruction files, Hermes runs any brain, and now Claude’s own desktop runs DeepSeek through Ollama. The moat keeps moving to whoever owns the memory and workflows underneath — the argument in the memory OS guide. Local wiring for the Hermes side is in Hermes + Ollama.
FAQ
Can Claude Desktop use DeepSeek models?
Yes — via Ollama’s official Claude Desktop support (25 August): toggle Claude on inside Ollama and choose any Ollama model, including DeepSeek’s open models, local or on Ollama’s cloud.
How hard is the setup?
Two steps: open Ollama, select Claude, toggle on. Ollama configures the third-party gateway for you.
Is my data sent to Anthropic or Ollama?
Per Ollama: no — telemetry is disabled by default and they state a strict Zero Data Retention policy across cloud and local.
Can I still use Anthropic’s models?
Yes — switch between Anthropic models and Ollama by toggling the feature, and turning it off restores your previous setup exactly.
Does this apply to the Claude Code CLI?
The announcement covers Claude Desktop. Treat CLI claims separately — this page doesn’t make them.
Local or Ollama cloud — which for DeepSeek?
Local for privacy and zero per-token cost if your hardware carries the model; Ollama’s cloud for the larger coding and research jobs.
The bottom line on Claude Code desktop DeepSeek setups
The Claude Code desktop DeepSeek pairing is now a two-step official reality: Ollama as a gateway inside Claude Desktop, any Ollama model behind Anthropic’s harness, one toggle back to stock. Run DeepSeek locally for the privacy, reach for Ollama’s cloud for the heavy jobs, and notice the bigger shift — the harness and the brain have officially stopped being a package deal.
About Julian Goldie
I run Goldie Agency, a 7-figure SEO agency, and teach this daily on a 400K+ subscriber YouTube channel. 240+ client projects on Upwork at a 100% job-success score, 10+ years through every major Google update. My systems are in the AI Profit Boardroom; my link building book is free here.
Related reading
Last updated August 2026. This is the living guide to Claude Code desktop DeepSeek — it gets updated as the tools change.