knowledge-work-plugins

1. Project Overview

knowledge-work-plugins is Anthropic's open-source collection of Claude plugins that turns Claude into a role-specific specialist — for sales, support, legal, finance, product management, and more — by packaging domain skills, slash commands, and tool connectors into a single installable unit.

2. Background & Positioning

Claude on its own is a general-purpose assistant. Most knowledge workers, however, operate inside a specific function with its own vocabulary, workflows, and toolchain — a support agent triages tickets in Intercom, a finance analyst reconciles ledgers in Snowflake, a product manager writes specs referencing Linear tickets. This repository was created to close that gap: each plugin teaches Claude how work is actually done in a given role, wires it up to the relevant systems via MCP connectors, and exposes both automatic domain expertise (skills) and explicit commands for common tasks.

Unlike generic prompt libraries or one-off automation scripts, knowledge-work-plugins is:

• File-based and buildless — plugins are markdown and JSON, no code or compilation step required.
• Designed for customization — every plugin ships as a starting point meant to be edited with your company's terminology, org structure, and tool stack.
• Dual-surface — the same plugins work in Claude Cowork (Anthropic's agentic desktop app) and in Claude Code.

3. Feature Categories

• 🗂️ Productivity — task and calendar management, personal workflow assistance. Connectors: Slack, Notion, Asana, Linear, Jira, Monday, ClickUp, Microsoft 365. Purpose: keep an individual's daily work organized across scattered tools.
• 💼 Sales — prospect research, call preparation, pipeline review, outreach drafting. Connectors: HubSpot, Close, Clay, ZoomInfo, Fireflies. Purpose: shorten the prep time before every sales interaction.
• 🎧 Customer Support — ticket triage, response drafting, escalation handling, knowledge base lookup. Connectors: Intercom, HubSpot, Guru, Jira. Purpose: help support teams resolve tickets faster and more consistently.
• 🧭 Product Management — spec writing, roadmap planning, research synthesis, competitive tracking. Connectors: Linear, Figma, Amplitude, Pendo. Purpose: turn scattered user feedback and data into decisions.
• 📣 Marketing — content drafting, campaign planning, brand enforcement, performance reporting. Connectors: Canva, Figma, Ahrefs, SimilarWeb, Klaviyo. Purpose: speed up content production while keeping brand consistency.
• ⚖️ Legal — contract review, NDA triage, compliance checks, risk assessment. Connectors: Box, Egnyte. Purpose: reduce the manual review burden on legal teams.
• 💰 Finance — journal entries, reconciliation, financial statement drafting, close management. Connectors: Snowflake, Databricks, BigQuery. Purpose: streamline recurring accounting cycles.
• 📊 Data — SQL querying, visualization, statistical analysis, dashboard building. Connectors: Snowflake, BigQuery, Hex. Purpose: let analysts go from question to chart faster.
• 🔎 Enterprise Search — cross-tool search across email, chat, docs, and wikis. Connectors: Slack, Notion, Guru, Jira, Asana. Purpose: answer "where is this documented" without manual hunting.
• 🧬 Bio Research — preclinical research support, genomics, target prioritization. Connectors: PubMed, BioRender, ChEMBL, Benchling. Purpose: accelerate literature review and early-stage research tasks.
• 🧩 Cowork Plugin Management — meta-plugin for creating and customizing new plugins for your own organization. Purpose: lower the barrier to building the next plugin.

4. Key Highlights

• Skills vs. commands separation — skills encode expertise Claude draws on automatically when relevant, while commands are explicit, user-triggered actions (e.g. /finance:reconciliation), giving predictable control alongside ambient intelligence.
• MCP-based connectors — every plugin's .mcp.json wires Claude to real external tools (Slack, Jira, Snowflake, HubSpot, and more) instead of relying on copy-pasted context.
• No build step — plugins are plain markdown and JSON, so anyone can read, fork, and edit them without a development environment.
• Marketplace distribution — plugins install through a simple marketplace mechanism in Claude Code, and directly from claude.com/plugins in Cowork.
• Built for customization, not just consumption — the README explicitly frames every plugin as a template: swap connectors, add company-specific terminology, and adjust workflows to match how your team actually works.
• Broad functional coverage — eleven plugins span sales, support, legal, finance, data, research, and more, making it one of the most complete open collections of role-based AI agent configurations.

5. Use Cases by Role

• General developers — use the Cowork Plugin Management plugin to scaffold new plugins, or study existing plugins as reference implementations for skills, commands, and MCP wiring.
• Data / research scientists — the Data plugin turns natural-language questions into SQL against Snowflake, BigQuery, or Hex, and the Bio Research plugin accelerates literature review and target prioritization workflows.
• Project / product managers — the Product Management plugin helps draft specs, plan roadmaps, and synthesize research pulled from Linear, Figma, Amplitude, and Pendo.
• Sales and support teams — the Sales and Customer Support plugins connect directly to CRM and ticketing systems to cut prep and response time.
• Finance and legal teams — the Finance plugin assists with reconciliation and close management; the Legal plugin handles contract review and NDA triage against documents stored in Box or Egnyte.

6. Getting Started

Find what you need — browse the plugin directories in the repository, or explore the marketplace listing:

https://github.com/anthropics/knowledge-work-plugins

Install / integrate — in Claude Code, add the marketplace and install a plugin:

claude plugin marketplace add anthropics/knowledge-work-plugins
claude plugin install sales@knowledge-work-plugins

In Claude Cowork, install directly from claude.com/plugins.

Contribute — fork the repository, edit or add a plugin (markdown and JSON only, no build step), and open a pull request:

git clone https://github.com/anthropics/knowledge-work-plugins

7. Project Structure

plugin-name/
├── .claude-plugin/plugin.json   # Plugin manifest (name, description, version)
├── .mcp.json                    # MCP server connections (external tool wiring)
├── commands/                    # Slash commands, explicitly invoked by users
└── skills/                      # Domain expertise, activated automatically

Each top-level directory in the repository (e.g. sales/, legal/, data/) is a self-contained plugin following this same layout, making it straightforward to copy one as a starting point for a new, organization-specific plugin.

8. Related Ecosystem

• Claude Cowork — Anthropic's agentic desktop application; the primary runtime for these plugins.
• Claude Code — Anthropic's CLI-based coding agent, which can also load these plugins via its marketplace mechanism.
• Model Context Protocol (MCP) — the open protocol used by each plugin's .mcp.json to connect Claude to external tools such as Slack, Jira, Snowflake, and HubSpot.

9. License

• ✅ Use, modify, and distribute the plugins, including for commercial purposes, under the Apache-2.0 license.
• ✅ Fork the repository and build proprietary, company-specific plugins on top of it.
• ❌ Do not remove existing copyright or license notices when redistributing.
• ℹ️ Apache-2.0 provides an explicit patent grant, which is worth noting if your organization plans to redistribute modified plugins internally or externally.

10. FAQ

Q: Do I need to write code to build a new plugin?
A: No. Plugins are composed of markdown (skills, commands) and JSON (manifest, MCP config) — no build step is required.

Q: Can I use these plugins outside Claude Cowork?
A: Yes. They are also compatible with Claude Code via claude plugin marketplace add anthropics/knowledge-work-plugins.

Q: How do I connect a plugin to my company's own tools instead of the defaults?
A: Edit the plugin's .mcp.json to point to your organization's MCP servers/connectors, and update the skill files with your own terminology and processes.

Q: Which plugin should I start with if my team doesn't fit any category exactly?
A: Start from the closest existing plugin (e.g. Productivity or Enterprise Search) and use the Cowork Plugin Management plugin to help adapt it to your workflow.

Q: How do I propose a new plugin or improvement?
A: Fork the repository, make your changes, and submit a pull request per the contributing guidelines.

11. Quick Links

• Repository: https://github.com/anthropics/knowledge-work-plugins
• Plugin marketplace (Cowork): https://claude.com/plugins/
• Issues: https://github.com/anthropics/knowledge-work-plugins/issues
• Pull requests: https://github.com/anthropics/knowledge-work-plugins/pulls

12. Summary

knowledge-work-plugins gives teams a ready-made, extensible library for turning Claude into a specialist across sales, support, legal, finance, product, data, and research functions. It is most valuable for organizations that want to deploy AI assistance quickly without building integrations from scratch, and for developers who want a clear, buildless reference for authoring their own Claude plugins.