SystemSculpt
Install free (opens Obsidian's plugin directory in a new tab)

DeepSeek in Obsidian: API Key, Ollama or Claude Code, Step by Step

Published 7 min read

On this page

To use DeepSeek in Obsidian, install a community plugin and connect it to DeepSeek in one of two ways. You can add a paid DeepSeek API key, or run a DeepSeek-R1 model on your own computer with Ollama. Obsidian has no AI of its own, so the plugin does the connecting.

The plugins worth starting with:

  • Copilot for chat with your vault. It takes a DeepSeek key or a local Ollama model.
  • Smart Composer for chat and one-click edits. It lists DeepSeek and Ollama as built-in providers.
  • Claudian for a coding agent in your vault. It supports DeepSeek Harness, DeepSeek's own agent, and Claude Code. Desktop only.

We make SystemSculpt, an Obsidian AI plugin that doesn't connect to DeepSeek, so this guide recommends other plugins. Facts checked November 3, 2026, against each plugin's README and DeepSeek's and Ollama's own docs.

Which setup fits

You wantUseCost
DeepSeek's newest models in chatCopilot or Smart Composer with a DeepSeek API keyPay per token from a prepaid balance
A model that runs on your computerCopilot or Smart Composer with Ollama and deepseek-r1Free, uses your computer's memory
An agent that reads and edits notesClaudian with DeepSeek Harness, or Claude Code pointed at DeepSeekPay per token

Route 1: DeepSeek's cloud API

Get a key

  1. Sign in at the DeepSeek Platform and create a key on the API keys page.
  2. Add money to your balance. DeepSeek's terms say you may need to prepay, and requests stop when the balance runs out.
  3. Check the current price first. DeepSeek charges per million tokens, with lower off-peak rates, and changes prices from time to time.

You need three values from DeepSeek's API docs:

  • Base URL (OpenAI format): https://api.deepseek.com
  • Base URL (Anthropic format): https://api.deepseek.com/anthropic
  • Model names: deepseek-flash and deepseek-v4-pro

DeepSeek V4 is current. deepseek-flash is served by DeepSeek-V4.1-Flash and supports images. deepseek-v4-pro is the larger model and costs more. Both have a 1M-token context window. Older guides use deepseek-chat and deepseek-reasoner, but DeepSeek's docs no longer list those names.

Copilot

Copilot's provider list loads inside the app, so its guide doesn't name every provider. The steps from its provider guide:

  1. Open Settings, Copilot, BYOK.
  2. Select Add a provider. Choose DeepSeek if it appears in the list. If not, choose Add a custom provider and enter the base URL https://api.deepseek.com.
  3. Paste your API key, then enter deepseek-flash or deepseek-v4-pro as the model.
  4. Select Test, then Save.

The model then shows up in Quick Chat, and in Copilot's OpenCode agent if OpenCode can route it.

Smart Composer

Smart Composer lists DeepSeek as a built-in provider, and its wiki links straight to DeepSeek's key page.

  1. Open Settings, Smart Composer, Providers, and add your DeepSeek API key.
  2. Pick a DeepSeek model in chat.
  3. If the model you want isn't listed, use Add custom model. Set Provider ID to your DeepSeek provider and Model to deepseek-flash.

Its README says the plugin is not under active development. The last release, 1.2.9, came out in January 2026, so its built-in model list may lag behind DeepSeek's names.

Claudian with DeepSeek Harness

DeepSeek Harness (dsh) is DeepSeek's open-source agent. Claudian, a community plugin, embeds coding agents in your vault, and DeepSeek Harness is one of them. That explains the "obsidian deepseek harness" search. The agent can read, write and search notes and run commands, with your vault as its working folder.

  1. Install Node.js, then set up DeepSeek Harness from its README. Its guide adds your key under Settings, Models.
  2. Install Claudian from Settings, Community plugins, Browse.
  3. In Settings, Claudian, Providers, open DeepSeek Harness and check that Claudian found it.

DeepSeek Harness is a developer preview. Its README warns of compatibility-breaking changes.

Claude Code on DeepSeek models

DeepSeek's API also speaks Anthropic's format, so Claude Code can run on DeepSeek models. DeepSeek's Claude Code guide sets these variables on Mac and Linux:

export ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic
export ANTHROPIC_AUTH_TOKEN=<your DeepSeek API Key>
export ANTHROPIC_MODEL=deepseek-flash[1m]
export ANTHROPIC_DEFAULT_OPUS_MODEL=deepseek-flash[1m]
export ANTHROPIC_DEFAULT_SONNET_MODEL=deepseek-flash[1m]
export ANTHROPIC_DEFAULT_HAIKU_MODEL=deepseek-flash
export CLAUDE_CODE_SUBAGENT_MODEL=deepseek-flash
export CLAUDE_CODE_EFFORT_LEVEL=max
export CLAUDE_CODE_AUTO_COMPACT_WINDOW=786432

Then run claude in your vault folder. Claude model names are mapped too: names starting with claude-opus go to deepseek-v4-pro, and claude-sonnet or claude-haiku go to deepseek-flash.

Inside Obsidian, Claudian's troubleshooting doc shows a Custom variables field for Claude Code under Settings, Claudian, Providers, Claude Code, Environment, which is where these variables would go. We haven't tested this setup, and we haven't tested SystemSculpt's Claude Code mode with DeepSeek, so this guide doesn't recommend it for that.

Route 2: DeepSeek on your computer with Ollama

Ollama runs models locally and listens at http://localhost:11434. The local DeepSeek family in Ollama's library is DeepSeek-R1, a reasoning model. The sizes:

TagDownloadBased on
deepseek-r1:1.5b1.1GBDeepSeek-R1-Distill-Qwen-1.5B
deepseek-r1:7b4.7GBDeepSeek-R1-Distill-Qwen-7B
deepseek-r1:8b (default)5.2GBDeepSeek-R1-0528-Qwen3-8B
deepseek-r1:14b9.0GBDeepSeek-R1-Distill-Qwen-14B
deepseek-r1:32b20GBDeepSeek-R1-Distill-Qwen-32B
deepseek-r1:70b43GBDeepSeek-R1-Distill-Llama-70B
deepseek-r1:671b404GBThe full DeepSeek-R1
  1. Install Ollama and open it.
  2. In a terminal, run ollama pull deepseek-r1 for the 8b model, or pick a smaller tag.
  3. Run ollama list and copy the exact name.

Ollama's library also has deepseek-v4.1-flash and deepseek-v4-pro, but only as :cloud tags. Those run on Ollama's cloud, not your computer.

Copilot: in Settings, Copilot, BYOK, select Add a provider and choose Ollama from the Self-host group. Copilot fills in http://localhost:11434/v1, and no key is needed. Enter your model name, select Test, then Save. If Test passes but chat can't send, edit the provider and turn on Enable CORS.

Smart Composer: add the Ollama provider with the base URL http://127.0.0.1:11434, then add a model with your Ollama model name, such as deepseek-r1:8b.

What leaves your computer

  • DeepSeek API, including Claude Code and DeepSeek Harness on it. Each prompt goes to DeepSeek with the notes the plugin attaches. DeepSeek's Open Platform terms say your personal information is handled under its privacy policy. That policy says DeepSeek collects, processes and stores personal data in the People's Republic of China. It also lists training its models among the uses.
  • Ollama with a local deepseek-r1 tag. The model runs on your machine, so notes and prompts don't go to an AI company.
  • Ollama :cloud tags. These send your prompts to Ollama's servers.

A plugin sends notes to whichever provider each chat uses. If you add a DeepSeek key next to a local model, check which one is selected.

Where SystemSculpt fits

SystemSculpt doesn't connect to DeepSeek, Ollama or any key you bring. It runs your installed Claude Code or Codex on desktop, free, with no SystemSculpt account. It also offers hosted SystemSculpt AI on desktop and mobile. Eligible new accounts can claim 500 one-time starter credits with no card. After that, Monthly is $9 a month and Yearly is $90 a year. See our guides to Claude Code in Obsidian and the best Obsidian AI plugins.

Questions

Is there a DeepSeek plugin for Obsidian? Not an official one. Use a general AI plugin: Copilot or Smart Composer for chat, or Claudian for an agent.

Does Obsidian Copilot support DeepSeek? Yes, as a BYOK provider. If DeepSeek isn't in the provider list, add a custom provider with https://api.deepseek.com.

What is DeepSeek Harness in Obsidian? It is DeepSeek's open-source coding agent. Claudian can run it in your vault. It is a developer preview.

Does DeepSeek V4 run in Obsidian? Yes, through the API as deepseek-flash or deepseek-v4-pro. It doesn't run locally through Ollama, where V4 is cloud only.

Can Claudian use DeepSeek? Yes, through DeepSeek Harness, which its README lists. Claude Code with DeepSeek's variables in Claudian's Custom variables field should also work, but we haven't tested it.

Does Text Generator support DeepSeek? Text Generator's README and docs don't cover DeepSeek setup, so we can't confirm the steps. Copilot and Smart Composer document theirs.

Is DeepSeek free in Obsidian? The API is paid from a prepaid balance. Running deepseek-r1 with Ollama is free.

Is the DeepSeek API private? Your notes go to DeepSeek, whose privacy policy says it stores data in China. For notes that stay on your computer, use Ollama.

Install SystemSculpt free from the Obsidian plugin directory to run Claude Code or Codex in your vault, or compare its plans for hosted AI.

Sources

Checked November 3, 2026.

TopicsObsidianObsidian AI

Try SystemSculpt inside your vault

SystemSculpt adds first-party AI chat, semantic search, transcription, and document workflows to Obsidian.

View pricing