dsh-plugin-subscriptions: A Subscription LLM Provider Plugin for DeepSeek Harness
v1ki/dsh-plugin-subscriptions
Use ChatGPT, Claude, and Grok subscriptions as LLM providers in DeepSeek Harness, no API keys, with OAuth login and usage tracking.
dsh-plugin-subscriptions is a subscription LLM provider plugin for DeepSeek Harness that lets you use your ChatGPT (Codex), Claude, Grok (X Premium), and GitHub Copilot subscriptions directly as model sources, eliminating API key management. Log in via OAuth or import existing credentials to seamlessly use these subscription models in the DSH web UI. The plugin offers live model catalogs, reasoning effort and speed controls, subscription usage monitoring, and extra search and image generation tools, maximizing the value of your subscriptions.
How to Install
dsh plugin --profile web add dsh-plugin-subscriptions- Category
- Models & Reasoning
- Platform
- DSH Plugin
- Author
- v1ki
- Distribution
- Plugin
dsh-plugin-subscriptions Key Features
dsh-plugin-subscriptions Repo Summary
What Does It Do?
It is a subscription-based LLM provider plugin for DeepSeek Harness (DSH) that lets you use your ChatGPT (Codex), Claude, Grok (X Premium), and GitHub Copilot subscriptions directly as model sources, with no API keys required. It solves the hassle of managing separate API keys for each model service by using OAuth login or importing existing credentials, so you can seamlessly use these subscription models inside the DSH web UI. Core capabilities include automatic token refresh, live model catalogs, reasoning effort selection, subscription usage display, and extra search and image generation tools.
Core Features
- Subscription login: Log in via OAuth in DSH's Settings → Subscriptions for Codex, Grok, and Copilot; Claude automatically imports credentials from Claude Code (macOS Keychain or ~/.claude/.credentials.json), so the Claude Code CLI is not required.
- Live model catalogs: After login, the session model picker shows real-time model lists from each provider (e.g., Codex from chatgpt.com/backend-api/codex/models, Grok from api.x.ai/v1/models), refreshed on login/logout.
- Reasoning effort and speed controls: Models with reasoning levels (e.g., Codex, Grok 4.6/4.5) get an Effort selector; Codex models support Standard/Fast speed toggle (service_tier: priority), also available via the /fast command.
- Subscription usage monitoring: Logged-in cards show usage percentage, progress bar, and reset time for each provider (Codex from wham/usage, Claude from api.anthropic.com/api/oauth/usage, Grok from cli-chat-proxy.grok.com/v1/billing); Copilot has no usage endpoint, so its card shows none.
- Extra tools: When the matching provider is enabled, registers x_search (Grok's hosted X search returning { answer, citations }), image_generate (generates gpt-image-2 via the Codex backend), and video_generate (plays generated clips inline).
How to Use This Plugin?
After enabling the plugin in DSH, go to Settings → Subscriptions in the web UI and click login for each provider: Codex and Grok use OAuth, Claude imports existing Claude Code credentials or falls back to browser OAuth, and Copilot uses the device flow (enter the shown code at github.com/login/device). Once logged in, the session model picker shows the available providers and their models; select a model to start chatting. To adjust reasoning effort or speed, use the Effort or Speed controls next to the model selector.
This page is an independent rewrite of the plugin's official README — for authoritative documentation and the latest changes, refer to the source: v1ki/dsh-plugin-subscriptions. The plugin is third-party code that runs on your machine once installed; inclusion does not imply endorsement — please review the source before installing.