Can You Use Knowledge Work Plugins with Claude Code and Claude Cowork?

Yes, the Knowledge Work Plugins are fully compatible with both Claude Code and Claude Cowork, requiring only standard installation via the Claude plugin marketplace.

The anthropics/knowledge-work-plugins repository provides 11 open-source plugins designed to enhance productivity across both the collaborative web interface (Claude Cowork) and the IDE-like environment (Claude Code). Because these plugins consist solely of static JSON and Markdown configuration files, they work interchangeably in either product without runtime compilation or platform-specific adaptations.

Architectural Design for Cross-Platform Compatibility

The plugins achieve universal compatibility through a declarative architecture that separates domain logic from execution. Both Claude Code and Claude Cowork share an identical plugin loading pipeline that reads static manifests and makes capabilities available at runtime.

Plugin Manifest Registration

Each plugin declares its identity through a plugin.json file located at .claude-plugin/plugin.json. This manifest specifies the plugin name, version, description, and author, which Claude's plugin loader reads to register the component.

{
  "name": "productivity",
  "version": "1.2.0",
  "description": "Manage tasks, plan your day, and build up memory of important context about your work.",
  "author": { "name": "Anthropic" }
}

Tool-Agnostic Connector Configuration

The .mcp.json file maps abstract connector placeholders to concrete Model Context Protocol (MCP) servers, making the plugins tool-agnostic. Placeholders like ~~calendar or ~~project tracker within skills and commands resolve to whatever calendar or project management service you configure, whether Microsoft 365, Google Workspace, or internal tools.

Skills and Commands Structure

The directory layout remains consistent across all 11 plugins:

plugin-name/
├── .claude-plugin/plugin.json   # Manifest for registration

├── .mcp.json                    # Connector wiring configuration

├── commands/                    # Slash command definitions (Markdown)

└── skills/                      # Domain expertise and workflows (Markdown)

Skills stored in skills/ directories encode best-practice workflows that Claude automatically activates when context matches. Slash commands defined in commands/ folders expose explicit triggers like /sales:call-prep that users invoke directly in chat interfaces.

Installing Knowledge Work Plugins

Installation requires only the Claude CLI, available in both Claude Code and Claude Cowork environments.

First, register the public marketplace that hosts the repository:

claude plugin marketplace add anthropics/knowledge-work-plugins

Then install any specific plugin (e.g., the sales plugin):

claude plugin install sales@knowledge-work-plugins

Once installed, the plugin's skills become auto-triggered based on conversation context, while slash commands appear in the command palette or UI.

Using Plugin Features in Both Interfaces

After installation, interaction patterns remain identical across both products.

To invoke a specific workflow, type the slash command in any Claude session:

/sales:call-prep

Claude executes the command defined in sales/commands/ and reaches out to connected tools via the placeholders configured in .mcp.json.

When writing skills, developers use connector placeholders to maintain tool flexibility:

To schedule a meeting, use the calendar connector:

1. Ask Claude to create an event with `~~calendar`.
2. Claude will call the MCP server you configured for Microsoft 365 (or any other calendar service).

Summary

• Knowledge Work Plugins from anthropics/knowledge-work-plugins work natively with both Claude Code and Claude Cowork.
• The static architecture (JSON manifests and Markdown files) requires no compilation, enabling instant portability between products.
• Installation follows a two-step process: adding the marketplace with claude plugin marketplace add, then installing specific plugins via claude plugin install.
• Tool-agnostic placeholders like ~~calendar in .mcp.json ensure skills work regardless of which specific SaaS tools your organization uses.
• Slash commands and skills activate automatically in both interfaces once the plugin manifest at .claude-plugin/plugin.json is loaded.

Frequently Asked Questions

Do I need separate installations for Claude Code and Claude Cowork?

No. Installing a plugin via claude plugin install makes it available across both environments simultaneously. The plugins reside in your user environment and load identically whether you access Claude through the web interface (Cowork) or the IDE-like Code interface.

What file formats do the Knowledge Work Plugins use?

The plugins use static JSON for configuration and Markdown for logic. The plugin.json manifest declares metadata, while .mcp.json handles connector mappings. Skills and commands are Markdown files stored in skills/ and commands/ directories, respectively. This declarative approach eliminates runtime dependencies.

Can I modify the connector placeholders to use my own tools?

Yes. Edit the .mcp.json file within any installed plugin to map abstract placeholders like ~~calendar to your specific MCP server endpoints. This configuration binds the plugin's generic workflows to your concrete tool stack without changing the underlying skill logic, as documented in the CONNECTORS.md file.

Are there any performance differences when using plugins in Claude Code versus Claude Cowork?

No. Since both products read the same static files (plugin.json, .mcp.json, and Markdown skills) through identical loading pipelines, performance characteristics remain equivalent. The plugins contain no compiled code or platform-specific binaries that would create divergence between the two interfaces.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}