knowledge-work-plugins

30.6K stars
Apache-2.0
Active this week
Last commit 2 days ago

Last verified Oct 20, 2026

Build role-specific AI plugins for knowledge workers in Claude, bundling skills, slash commands, and connectors for 11 job functions.

Visit website GitHub30.6K

Read the full breakdownWho it is for, how it works, trade-offs, the quick start and the FAQ

who it's for

Who knowledge-work-plugins is for#

Sales teams preparing for customer calls

The sales plugin connects to HubSpot, Close, Clay, ZoomInfo, and Fireflies, giving Claude context to research prospects, summarize prior call notes, surface pipeline status, and draft outreach. The /sales:call-prep command ties these into one workflow a rep can trigger before any meeting.

Skip if:

Your sales team runs on a deeply customized Salesforce instance. The plugin's out-of-box connectors cover HubSpot, Close, and Clay; a Salesforce setup needs custom connector work before the plugin adds value.

Product managers writing specs and roadmap updates

The product-management plugin connects to Linear, Jira, Asana, Figma, Amplitude, Pendo, and Intercom. Claude can pull user research, summarize feedback, draft a PRD in the team's format, and produce stakeholder updates at the right level of detail for each audience.

Skip if:

Your PM workflow centers on a tool not in the connector list (Productboard, Aha). You can add connectors manually, but there is setup work before the plugin adds full value.

Finance teams running close and reconciliation

The finance plugin integrates with Snowflake, Databricks, BigQuery, Slack, and Microsoft 365. Claude can prep journal entries, reconcile accounts, generate variance analysis, and support audit prep from inside your data infrastructure via the /finance:reconciliation command.

Skip if:

Your finance data lives in a warehouse not on the connector list. The plugin depends on Claude querying your data directly; a different warehouse needs a custom MCP connector first.

Organizations building plugins for uncovered roles

The cowork-plugin-management plugin is the meta-layer for extending the library. Use it to build plugins for HR, recruiting, IT support, or executive assistants. Creating a new plugin means authoring domain knowledge and connector config as markdown and JSON, not writing code.

Skip if:

Your use case is deeply technical (developer tooling, DevOps workflows) rather than a knowledge-worker role. Claude Code's native skills system is better suited to code-centric work.

the problem

The problem it solves#

Knowledge workers in specialist roles get a general-purpose AI assistant when their role demands one that already knows the workflow. A sales rep needs Claude to know their CRM, call prep format, and outreach voice. A product manager needs it to understand the spec template engineering expects, the stakeholder update cadence, and the roadmap structure their team uses. A finance analyst needs it to speak the close calendar and reconciliation process.

General-purpose prompting covers basic tasks but breaks down on the recurring workflows that define a role. Each specialist ends up manually injecting context into every session, and that context disappears when the session ends. Teams that want consistent AI behavior across functions have no way to bake shared context into the assistant without writing custom infrastructure or accepting that the AI's knowledge resets daily.

how knowledge-work-plugins solves it

How it solves it#

11 job-function plugins included

Ships plugins for productivity, sales, customer support, product management, marketing, legal, finance, data analysis, enterprise search, bio-research, and plugin management. Each covers the recurring workflows and tool integrations its function depends on. Install any via two commands in Claude Code or one click in Claude Cowork.

Markdown and JSON structure with no build steps

Every plugin is plain text: skill files are markdown encoding domain expertise Claude draws on automatically, connectors are .mcp.json files pointing at MCP servers, and commands are markdown files defining slash triggers like /sales:call-prep. No code, no compiled artifacts, no build pipeline. Customizing a plugin means editing text files and reinstalling.

MCP connector library wired per job function

Each plugin ships a .mcp.json connecting Claude to the tools that role depends on. Sales connects to HubSpot, Close, Clay, ZoomInfo, and Fireflies. Product management connects to Linear, Figma, Amplitude, and Pendo. Finance connects to Snowflake, Databricks, and BigQuery. Replace or extend any connector by editing the JSON.

Named slash commands for repeatable workflows

Beyond the skills Claude draws on automatically, each plugin exposes commands for explicit workflows: /sales:call-prep, /data:write-query, /product-management:write-spec, /finance:reconciliation. These give teams a consistent, namespaced entry point for the tasks they run most often, triggerable from inside any Claude session.

Built-in plugin authoring for uncovered roles

The cowork-plugin-management plugin lets you create new plugins or customize existing ones from inside Claude. Every plugin follows the same file structure, so building one for a role not yet covered (HR, recruiting, IT support) means writing markdown and JSON, not code. No plugin SDK or developer environment required.

strengths · trade-offs

Strengths and trade-offs#

Strengths

Apache-2.0 with full inspection rightsEvery skill file, command definition, and connector config is readable text under Apache-2.0. Unlike Microsoft Copilot or Salesforce Einstein, where the AI's embedded behaviors are opaque, you can open any plugin file, read what Claude knows about a workflow, and edit it directly. Apache-2.0 permits internal commercial use and modification with no restriction.
No lock-in on connector or tool stackEach plugin's .mcp.json is a config file pointing at MCP servers. Point it at your CRM instead of the default, add connectors the plugin does not include, or replace any tool in the list. You are not locked to the connectors Anthropic chose for a generic starting point.
24,000+ stars and actively maintained since January 2026Over 24,000 GitHub stars and 2,900+ forks since the January 2026 launch, with daily pushes. The community size means new plugins for uncovered roles surface quickly and connector inconsistencies are caught by a large user base before they become widespread.
Works across Cowork and Claude CodePlugins install identically in Claude Cowork (one click) and Claude Code (two CLI commands). Teams running both share the same plugin library, with the same skill files and slash commands available in both environments. No separate configuration per deployment path.

Trade-offs

-Requires a paid Claude subscriptionThese plugins run only inside Claude's plugin runtime (Cowork or Claude Code). Teams not using Claude commercially cannot access the plugin system. There is no way to run these plugins on a self-hosted open source LLM or any alternative AI provider.
-79 open issues on a fast-moving codebaseWith 79 open issues and active development since January 2026, connector inconsistencies and edge cases exist. Configurations for less-common tools (Pendo, Hex, Egnyte) may need adjustment for specific environments. Check the issue tracker before writing a custom connector to avoid duplicating work already in progress.

versus alternatives

knowledge-work-plugins vs alternatives#

Knowledge Work Plugins vs Microsoft Copilot

Microsoft Copilot is the closest commercial analog: role-specific AI behaviors embedded in a productivity platform, with specialized versions for sales (Dynamics 365 Copilot), customer service (Copilot for Service), and general knowledge work (Copilot for Microsoft 365).

Feature	Knowledge Work Plugins	Microsoft Copilot
License	Apache-2.0	Proprietary
Platform	Claude Cowork / Claude Code	Microsoft 365 / Dynamics 365
Inspect AI behavior	Yes (plain text files)	No
Customize behaviors	Yes (edit skill files)	Limited (Copilot Studio)
Connector standard	MCP (open)	Microsoft Graph API (proprietary)
Pricing	Included with Claude subscription	$30+/user/month add-on

Knowledge Work Plugins is the better choice when your team is already on Claude, you need to customize AI behavior beyond what a no-code builder supports, or you want to inspect and version-control the exact expertise baked into your assistant. Every skill file is readable text a team lead can open and edit without vendor involvement.

Microsoft Copilot is the stronger choice when your organization runs deeply on Microsoft 365. Its Graph API integration pulls from Teams, SharePoint, Outlook, and OneDrive without connector configuration. If your source of truth is Microsoft tools and you are not already on Claude commercially, Copilot has less setup friction.

Knowledge Work Plugins vs Salesforce Einstein

Salesforce Einstein overlaps most directly with the sales and customer-support plugins. Einstein is embedded in Sales Cloud and Service Cloud with no separate install.

Feature	Knowledge Work Plugins	Salesforce Einstein
License	Apache-2.0	Proprietary
CRM dependency	Optional (HubSpot, Close, others)	Requires Salesforce CRM
Inspect AI behavior	Yes (skill files)	No
Self-hosting	Inside your Claude deployment	Salesforce-managed only
Pricing	Included with Claude subscription	Bundled in Salesforce pricing

For teams not on Salesforce, Knowledge Work Plugins reaches the same sales-assist and support-assist functionality without a CRM migration. If your stack is HubSpot, Close, or Clay, the sales plugin connects natively. Einstein is the better fit only when Salesforce is already your CRM and you want AI that surfaces Salesforce data without additional configuration.

install · quick start

Quick start#

bash

Install plugins from the Claude Code CLI after adding the repository as a marketplace source, or clone it locally to browse and customize.
```bash
git clone https://github.com/anthropics/knowledge-work-plugins
claude plugin marketplace add anthropics/knowledge-work-plugins
claude plugin install sales@knowledge-work-plugins
```

tech stack · detected from GitHub

What it's built on#

Languages: Python

frequently asked

FAQ#

Do I need a Claude subscription to use these plugins?

Yes. These plugins run inside Claude's plugin runtime, available in Claude Cowork (a team subscription) and Claude Code (an API subscription). There is no standalone version; the plugins extend Claude's behavior and cannot run without it.

Can I use these plugins with OpenAI, Gemini, or other AI providers?

No. The skill files, command format, and MCP connector structure are specific to Claude's plugin runtime. The plugins depend on Claude's ability to interpret skill markdown and route slash commands. Porting them to another provider would require rebuilding that runtime layer.

Can I add my company's internal tools as connectors?

Yes. Each plugin's .mcp.json lists the MCP servers it connects to. Add an entry pointing at your internal tool's MCP server and reinstall the plugin. Your tool needs an MCP-compatible server or adapter as a prerequisite. The repository README covers swapping and extending connectors.

What license applies to plugins I build on top of this repository?

The repository is Apache-2.0 licensed, which permits forking, modifying, and distributing your own plugins for internal organizational use without opening your customizations. If you redistribute modified versions publicly, you must preserve the Apache-2.0 license notice and attribution.

How do I keep plugins up to date as Anthropic releases changes?

Run the marketplace add command again to pull the latest version, then reinstall the specific plugins you use. For customized forks, pull upstream changes from the repository and merge them with your modifications manually. The plugin files are plain text, so merges are straightforward.

also worth a look

Similar open-source tools#

FckSignups

InCustomer Support

Open-source tools that work instantly, no signup required

4,511 GitHub stars
Licence GPL-3.0

SuiteCRM

Alternative toSalesforce Sales Cloud

Self-hosted CRM with leads, quotes, and workflow automation.

5,805 GitHub stars
Licence AGPL-3.0

crm

Alternative toHubSpot CRM

Self-hostable CRM with unlimited users and Kanban deal tracking

3,837 GitHub stars
Licence AGPL-3.0

DeskcommCRM

Alternative toHubSpot Sales Hub

AI agents for WhatsApp sales, fully audited, no per-user pricing

4,751 GitHub stars
Licence MIT

fmt

InCustomer Support

Fast, type-safe C++ formatting that replaces printf and iostreams

25,884 GitHub stars
Licence MIT

agency-agents

InAI & Machine Learning

Expert AI agent personalities for every workflow

162,144 GitHub stars
Licence MIT

by category

12,712 GitHub stars
Licence AGPL-3.0
#Email

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

4,904 GitHub stars
Licence Apache-2.0
#AI Agents

Omnara

Alternative toClaude Managed Agents

Open-source agent deployment API. Self-host or use Omnara Cloud.

2,957 GitHub stars
Licence Apache-2.0
#AI Agents

Feature your project

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Knowledge Work Plugins vs Microsoft Copilot

Feature	Knowledge Work Plugins	Microsoft Copilot
License	Apache-2.0	Proprietary
Platform	Claude Cowork / Claude Code	Microsoft 365 / Dynamics 365
Inspect AI behavior	Yes (plain text files)	No
Customize behaviors	Yes (edit skill files)	Limited (Copilot Studio)
Connector standard	MCP (open)	Microsoft Graph API (proprietary)
Pricing	Included with Claude subscription	$30+/user/month add-on

Knowledge Work Plugins vs Salesforce Einstein

Salesforce Einstein overlaps most directly with the sales and customer-support plugins. Einstein is embedded in Sales Cloud and Service Cloud with no separate install.

Feature	Knowledge Work Plugins	Salesforce Einstein
License	Apache-2.0	Proprietary
CRM dependency	Optional (HubSpot, Close, others)	Requires Salesforce CRM
Inspect AI behavior	Yes (skill files)	No
Self-hosting	Inside your Claude deployment	Salesforce-managed only
Pricing	Included with Claude subscription	Bundled in Salesforce pricing