dsh-agent-studio
An open-source visual AI Agent configuration plugin that supports graphical assembly and management of prompts, toolsets, MCP, Skills, and sub-Agents. It is designed to address context contamination issues that arise during Agent development by providing a structured, modular visual interface for composing agent pipelines.
Core Facts
Timeline (last 90 days)
dsh-agent-studio (DSH plugin) is an open-source plugin for visually configuring AI Agent prompts, tools, MCP, Skills, and fallback models.
dsh-agent-studio was developed and open-sourced by a Bilibili content creator.
The plugin allows the creation of clean Agents with no preset context, building capability combinations from scratch.
The plugin supports forming Agent teams, where each sub-Agent can have its own model, reasoning intensity, and fallback model configured individually.
The plugin's configuration files support global and local import/export, making it easy to reuse across projects.
In the plugin, multiple prompts are concatenated in a set order and sent to the AI together.
In the plugin, checking a tool means authorizing the AI to autonomously invoke those capabilities, providing explicit boundary control.
All Facts (7)
In the plugin, checking a tool means authorizing the AI to autonomously invoke those capabilities, providing explicit boundary control.
65%Unverifieddsh-agent-studio was developed and open-sourced by a Bilibili content creator.
50%UnverifiedThe plugin allows the creation of clean Agents with no preset context, building capability combinations from scratch.
50%Unverifieddsh-agent-studio (DSH plugin) is an open-source plugin for visually configuring AI Agent prompts, tools, MCP, Skills, and fallback models.
50%UnverifiedThe plugin's configuration files support global and local import/export, making it easy to reuse across projects.
50%UnverifiedIn the plugin, multiple prompts are concatenated in a set order and sent to the AI together.
50%UnverifiedThe plugin supports forming Agent teams, where each sub-Agent can have its own model, reasoning intensity, and fallback model configured individually.
50%