dsh-agent-router
Run the following command in DeepSeek Harness:
dsh plugin install peterwangze/dsh-agent-router
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install peterwangze/dsh-agent-router in your terminal to install the plugin; source code is at https://github.com/peterwangze/dsh-agent-router . Restart DeepSeek Harness after installation.
About this plugin
A DSH main agent excels at long-form reasoning and code generation, yet falls short when faced with screenshot analysis, image generation, speech transcription, translation, or multi-step subtasks. Manually switching models, juggling multiple API accounts, and configuring a separate subagent for each capability is cumbersome, with settings scattered across the system and model reselection needed for every new session.
dsh-agent-router delegates specialized tasks to specialized agents: configure default models per DSH preset for both the main agent and subagents with instant visibility on session open; define custom agents for vision, image generation, translation, and speech with automatic capability-tag routing, each bound to its own provider and model; log in to a ChatGPT subscription via official OAuth and use subscribed models directly as the main model, or drive image generation with the subscription. It also supports headless CLI subagents (Codex, Claude Code, Gemini CLI), multi-account pools with automatic failover, and real-time usage statistics at both agent and account levels.
Built for DSH users who want to expand the capability boundary of their main agent: those who need to mix multiple models in a single session, manage several provider accounts, or weave image, speech, and translation capabilities seamlessly into the conversation flow. Zero-config integration—install, restart, and go.
Screenshots
Use Cases
- Paste a screenshot in chat and it is automatically routed to a vision agent while the main model never touches image bytes
- Delegate multi-step coding or research tasks to headless CLI subagents like Codex or Claude Code in the workspace
- Assign different default models per DSH preset so new sessions land on the right model without manual switching
Best For
- Developers who want to extend the DSH main agent and mix multiple models in a single session
- Teams managing multiple provider accounts with automatic account selection and failover
- Users who need to weave image understanding, speech transcription, and translation seamlessly into conversation flow
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.


