AI Agent Hub
Back to plugins
🖥️

dsh-agent-manager

Client Updated 2026.08.31

Run the following command in DeepSeek Harness:

dsh plugin install hrhgit/dsh-agent-manager

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install hrhgit/dsh-agent-manager in DeepSeek Harness to add this plugin; source repository: https://github.com/hrhgit/dsh-agent-manager

About this plugin

Agent presets in DeepSeek Harness often declare dozens of tool rows, but until now there was no built-in UI to see at a glance which tools are actually enabled, which are unavailable, and which are locked down by the host. dsh-agent-manager closes that gap by adding an Agent Manager section to the Harness settings page, turning opaque JSON configuration into a clear, interactive roster.

The plugin surfaces only the rows that explicitly opt in via agentTool.label for the currently selected preset and annotates each with a live status indicator: host-controlled, not mounted, available, unavailable, or stale. In locally authored presets, users can toggle the disabled field directly from the UI. Shipped and other system-owned presets stay read-only until you copy them into a user preset. Policy edits apply to subsequent sessions; a running session keeps the tool composition it started with, so there are no mid-session surprises.

If you maintain several agent presets and need a quick way to confirm tool visibility and enable-disable state without spelunking through JSON, this plugin is for you. It intentionally avoids touching model providers or model metadata, delegating roster discovery, validation, locking, atomic writes, and runtime status entirely to the Host's ctx.agentPresets service. On the client side, the only thing persisted in local browser storage is the index of the currently selected preset, keeping the plugin lightweight and its responsibilities tightly scoped.

Use Cases

  • Check the enable, disable, or unavailable status of each tool in the current agent preset from the settings page
  • Toggle tool disabled flags directly in the UI for locally authored presets without manually editing JSON
  • Distinguish between read-only system presets and editable user presets to avoid modifying protected configs
  • Confirm that a policy change applies only to subsequent sessions, leaving the current session untouched

Best For

  • Platform developers maintaining multiple agent presets in DeepSeek Harness
  • Preset authors who need to verify tool visibility and enable/disable state
  • End users who prefer a visual interface over manual JSON editing for tool policy management