The project's README from github.com/Zhiyuan-Fan/Awesome-DeepSeek-Harness-Plugins. As one markdown file: README.md.
A concise, daily-curated directory of public plugins and extensions for DeepSeek Harness (DSH), the open-source DeepSeek agent harness. Explore tools, skills, model providers, memory, automation, runtimes, desktop clients, browser integrations, and developer utilities.
Every entry links to its canonical GitHub repository, current star count, and independently verified references.
Inclusion: public, useful, maintained, and clearly built for DSH. Review a plugin's code and permissions before installing it.
| Project | Description | Stars | Referenced by |
|---|---|---|---|
| DeepSeek Harness | A developer-preview agent harness where every capability is a swappable, composable plugin. | 1 page |
|
| DeepSeek Harness Handbook | Provides source-backed operator guides for DeepSeek Harness architecture, plugins, MCP, sessions, sandboxing, evaluation, and troubleshooting. | 1 page
|
Taxonomy: Models & Providers · Tools & Skills · Sessions & Storage · Loops & Scheduling · Runtime & Sandboxes · UI & Clients · Integrations · Developer & Operations
Browse: Models & Providers · Tools & Skills · Sessions & Storage · Loops & Scheduling · Runtime & Sandboxes · UI & Clients · Integrations · Developer & Operations
The taxonomy follows the capability seams defined by the official Harness overview.
Projects in each section are ranked by GitHub stars, then by verified reference count.
| Project | Description | Stars | Referenced by |
|---|---|---|---|
| dsh-routing-suite | Bundles a runtime injector with an experimental preset that routes DSH sessions among task-aware reasoning modes. | 0 pagesNo verified references yet. |
|
| dsh-plugin-subscriptions | Adds DSH model providers for eligible ChatGPT, Claude, Grok, and GitHub Copilot subscriptions through OAuth, device login, or existing local credentials. | 0 pagesNo verified references yet. |
|
| dsh-commandcode-provider | Adds an unofficial Command Code provider with browser sign-in, live model discovery, plan filtering, account rotation, image input, and Web search. | 0 pagesNo verified references yet. |
|
| dsh-codex-connect | Adds ChatGPT OAuth model access and optional GPT Image generation to DSH, with diagnostics and session recovery through native Harness services. | 0 pagesNo verified references yet. |
|
| dockyard-dsh | Connects macOS DSH profiles to official provider CLI and OAuth sessions for account pooling, model discovery, quota views, and provider-native requests. | 0 pagesNo verified references yet. |
|
| rapid-mlx-dsh-provider | Adds a configurable OpenAI-compatible Rapid-MLX provider with model discovery, advertised context limits, and local-server routing. | 0 pagesNo verified references yet. |
|
| dsh-codex-subscription | Adds ChatGPT or Codex subscription OAuth, models, quotas, search, speed controls, usage resets, and image generation or editing to DSH. | 0 pagesNo verified references yet. |
|
| dsh-agy-link | Connects Google Antigravity CLI models to DSH with streaming responses, tool activity, usage reporting, and Web OAuth sign-in. | 0 pagesNo verified references yet. |
|
| dsh-workbuddy-connect | Connects models from a locally installed and signed-in WorkBuddy desktop app to DSH Web, Desktop, or TUI with account status and remaining-credit views. | 0 pagesNo verified references yet. |
|
| deepseek-harness-model-config | Adds a Web editor for custom model endpoints, protocols, headers, credentials, capability metadata, retries, and models.dev enrichment. | 0 pagesNo verified references yet. |
|
| opencode2dsh | Adds OpenCode Zen models to DSH through a native provider with live catalog discovery, offline fallback, and classified upstream errors. | 0 pagesNo verified references yet. |
|
| dsh-llm-fallbacks | Adds role-based model retries, ordered provider fallbacks, and time-based routing for DSH Web and terminal profiles. | 0 pagesNo verified references yet. |
|
| dsh-advisor | Uses a separate configured model to review DSH turns and inject severity-ranked advisory notes without approving or rejecting the primary agent's actions. | 0 pagesNo verified references yet. |
|
| dsh-plugin-subagent-director | Assigns provider and model choices plus reusable role templates to individual DSH subagents. | 0 pagesNo verified references yet. |
|
| dsh-codex-auth | Adds a DSH Codex capability bundle that reuses the official Codex login for models, search, usage, and image tools. | 0 pagesNo verified references yet. |
|
| dsh-router-flash | Installs a versioned OpenCode Go agent preset with a Flash-oriented persona, routing settings, and update checks. | 0 pagesNo verified references yet. |
|
| dsh-openai-codex-auth | Adds OpenAI Codex OAuth sign-in and usage status to DeepSeek Harness. | 0 pagesNo verified references yet. |
|
| dsh-ocgo-lite | Shows OpenCode Go account quota plus DSH-only token and cost usage in a persistent Web bar and model tool. | 0 pagesNo verified references yet. |
|
| dsh-local-ai | Adds an Ollama provider with model management, health checks, rule-based local routing, and cloud fallback. | 0 pagesNo verified references yet. |
|
| dsh-deliberation-presets | Provides three experimental DeepSeek Harness presets for separating deliberation from tool execution while preserving tool continuity. | 0 pagesNo verified references yet. |
|
| DSH-Brain-Bridge | Connects DSH agents to ChatGPT Web as an external reasoning provider through a journaled MCP gateway with worker failover. | 0 pagesNo verified references yet. |
|
| dsh-translate | Maps sampling parameters across 11 model vendors and deterministically repairs malformed JSON tool results. | 0 pagesNo verified references yet. |
|
| dsh-model-meta-autofill | Fills missing custom-model context, output, display-name, and modality metadata from a bundled and refreshable models.dev index. | 0 pagesNo verified references yet. |
|
| dsh-role-router | Routes planner and subagent roles to configured models and exposes the active routing summary in DSH Web. | 0 pagesNo verified references yet. |
|
| dsh-custom-provider-settings | Adds headers, image-capability declarations, and reasoning-level controls for user-defined model providers without modifying Harness source. | 0 pagesNo verified references yet. |
|
| dsh-custom-provider-reasoning | Adds reasoning-level metadata and missing token limits to eligible custom providers while preserving explicit model settings and protocol-specific compatibility. | 0 pagesNo verified references yet. |
|
| dsh-thinking-levels-settings | Adds a settings page for configuring per-model reasoning levels on custom pi-ai providers through the native settings wire. | 0 pagesNo verified references yet. |
|
| dsh-everything-oauth | Imports existing local OAuth sessions for Codex, Grok, Claude, OpenCode, and related tools. | 0 pagesNo verified references yet. |
|
| dsh-llm-ollama | Adds Ollama Cloud model discovery and chat through a native DSH provider, plus Ollama-native web search and page fetching. | 0 pagesNo verified references yet. |
|
| dsh-plugin-vision-bridge | Routes image attachments through a configured OpenAI-compatible vision model and passes durable textual descriptions to text-only DeepSeek routes. | 0 pagesNo verified references yet. |
|
| dsh-cot-summerization | Intercepts raw reasoning streams and replaces them with incremental summaries from a configured auxiliary model. | 0 pagesNo verified references yet. |
|
| dsh-models-radar | Displays public benchmark snapshots, model comparisons, cost views, and the selected model's DeepSWE score in DSH, with per-result harness attribution. | 0 pagesNo verified references yet. |
|
| dsh-vision | Gives text-only routes image understanding by substituting cached descriptions from an available vision model or system OCR for provider calls. | 0 pagesNo verified references yet. |
|
| dsh-model-hub | Unifies provider sign-in, model metadata, default selection, composer routing, and the /model command behind a loopback-only DSH service. | 0 pagesNo verified references yet. |
|
| dsh-plugin-model-capability | Adds bilingual DSH controls for model reasoning levels, context and output limits, input modalities, and gateway compatibility presets. | 0 pagesNo verified references yet. |
|
| dsh-qwen38-compaction-fix | Adjusts matched Qwen3.8 requests during DSH context compaction so reasoning does not consume the summary output budget, while leaving normal turns unchanged. | 0 pagesNo verified references yet. |
|
| dsh-llm-codex | Adds a ChatGPT subscription adapter to the DSH model seam with PKCE or device OAuth, token refresh, and Codex Responses forwarding. | 0 pagesNo verified references yet. |
|
| dsh-model-memory | Adds custom-provider reasoning-level controls, retry policies, and persistent per-mod |