Plugins

Bundle skills, commands, sub-agents, hooks, and MCP servers into reusable plugin packages.

Field reference: plugins — every field, type, default, and tier for this page.

Plugins let you package a set of capabilities — skills, commands, sub-agents, hooks, and MCP servers — into a single reusable unit. You can install a plugin on multiple agents or share it with others.

Today the engine consumes two of a plugin’s components at chat time: its MCP servers (connected as tools) and its Commands (a message of the form [COMMAND:/name] expands that command’s prompt template before the LLM call). Skills, sub-agents, and hooks declared inside a plugin are saved here but not yet active during conversations.

What a plugin contains

Component	Description
MCP servers	External MCP-compatible servers the agent connects to when the plugin is active.
Skills	Skill stubs (name + description) bundled with the plugin. Saved but not yet active at chat time.
Commands	Slash-style commands (name, description, prompt template). Active at chat time — a [COMMAND:/name] message expands the command’s prompt template before the LLM call.
Sub-Agents	Agent definitions (name, description, system prompt, tool restrictions) the plugin contributes. Saved but not yet active at chat time.
Hooks	Event hooks declared by the plugin (event, command, description). Saved but not yet active at chat time.

Each plugin also has a Name, Description, and Version (must look like 1.0 or 1.2.3, optionally with a prerelease/build suffix).

Adding a plugin

1. Open your agent and go to Plugins in the sidebar.
2. Click Add Plugin to create an empty plugin in-place, or Add Plugin From Directory to install one from the plugin directory.
3. Fill in the name, description, and version.
4. Expand the + Skills, + Commands, + Agents, + Hooks, or + MCP Servers sections to add components.
5. Click Save Changes.

Adding components

Inside each expandable section, click the corresponding Add … button to add a row. Each row has its own fields (e.g. an MCP server takes a name, URL, and description; a command takes a name, description, and prompt template).

Sub-agent tool restrictions accept a comma-separated list of tool names; the input auto-completes against the agent’s available tools.

Examples

Hello world — install from the directory. Click Add Plugin From Directory, pick a plugin, and click Install. This adds an entry to your plugin list, but its component arrays (MCP Servers, Commands, Skills, Sub-Agents, Hooks) start empty — nothing auto-activates. Open the new plugin card and populate its MCP Servers and Commands yourself, following the plugin’s README.

Medium — a reusable team toolkit. Build a plugin named support-toolkit that bundles the MCP server for your help-desk, two skills (ticket lookup, canned-response formatter), and a /escalate command. Save it once and install it into every support agent you run, so they all share the same tooling without re-entering it.

Complex — a packaged sub-agent system. Build a plugin that contributes a sub-agent (e.g. a “researcher” with a restricted tool set), the MCP servers that sub-agent needs, hooks that log its activity, and a command that invokes it. The plugin becomes a self-contained extension — a whole specialist capability, its tools, its governance, and its entry point — that travels as one installable unit across your fleet.

Plugin order

Plugins appear in the list in the order they were added. The engine merges their MCP servers in that order; if two plugins provide servers with the same name, the first one wins.

Enabling and disabling a plugin

Every plugin carries an enabled flag (default true). When it is false, the engine skips that plugin’s MCP servers and commands at chat time — the plugin’s configuration stays saved but does nothing.

There is no on/off control for this flag in the dashboard; it can only be set through the Management MCP. Because the plugins section is an array, config.set writes the whole list, so include every plugin you want to keep and set enabled: false on the one you are taking out of service:

{
  "tool": "config.set",
  "owner_id": "<your owner_id>",
  "agent_id": "<your agent_id>",
  "section": "plugins",
  "data": [
    {
      "name": "support-toolkit",
      "version": "1.0.0",
      "enabled": false,
      "mcp_servers": [],
      "commands": []
    }
  ]
}

Manage via the Management MCP

This page can be configured programmatically through the Management MCP — useful for AI agents and CI pipelines.

Endpoint	https://manage.agentheya.com/mcp
Tool	config.set
Section	plugins
Schema (HTTP)	GET /api/manage-mcp/schema/plugins

Example:

{
  "tool": "config.set",
  "owner_id": "<your owner_id>",
  "agent_id": "<your agent_id>",
  "section": "plugins",
  "data": { /* see schema for accepted fields */ }
}

Partial updates are supported — only the fields you include are changed. Call config.get to read the current values and schema.get (or the HTTP schema URL above) to see the full field list.