Claude Cowork is an AI agent that works like a colleague and completes tasks on its own — e.g. building presentations or doing bookkeeping. Cowork uses the same agentic architecture as Claude Code but is now directly accessible inside Claude Desktop, without opening a coding terminal. Instead of reacting to individual requests one at a time, Claude can take on complex, multi-step tasks and run them on your behalf.
With Cowork you describe a desired outcome, focus on something else, and come back later to finished work — formatted documents, organised files, summarised research, and more. With Scheduled Tasks, Claude can run work automatically for you — something that's not possible in regular chats outside of Cowork.
Cowork is available as a research preview on paid plans (Pro, Max, Team, Enterprise) on:
Claude plugins are specialised extensions for the Claude Desktop app (Chat, Cowork, Code) that bring domain-specific knowledge and structured workflows. Every plugin contains skills (auto-activated expertise) and commands (explicit slash commands for structured flows).
Plugins work standalone. All commands work without external connections. Optionally MCP connectors (Slack, Notion, Jira, GitHub, etc.) can be attached to extend capabilities.
The following five plugins are ordered by priority and form the core set for digital product development and productive teams.
| Priority | Plugin | Version | Skills | Commands | Primary value |
|---|---|---|---|---|---|
| 1 | Engineering | 1.1.0 | 6 | 6 | Code reviews, architecture, incident response |
| 2 | Design | 1.1.0 | 6 | 6 | UX critique, accessibility, design systems |
| 3 | Product Management | 1.1.0 | 6 | 6 | PRDs, roadmaps, stakeholder communication |
| 4 | Marketing | 1.1.0 | 5 | 7 | Content, campaigns, performance analytics |
| 5 | Brand Voice | 1.0.0 | 3 | 3 | Brand guidelines, brand discovery, enforcement |
Version: 1.1.0
Code Review — checks code for bugs, security holes, performance, and maintainability. Detects SQL injection, XSS, CSRF, N+1 queries, race conditions, and edge cases.
Documentation — produces technical docs: READMEs, API docs, architecture docs, runbooks, and onboarding guides.
Incident Response — triage and manage production incidents. Framework with severity classification (SEV1–4), response phases, and blameless postmortem format.
System Design — design systems, services, and architectures. Framework: requirements gathering, high-level design, deep dive, scaling/reliability, trade-off analysis.
Tech Debt — identifies, categorises, and prioritises technical debt.
Testing Strategy — designs test strategies and test plans. Test pyramid: unit tests, integration tests, E2E tests.
| Command | Description |
|---|---|
/standup | Generate a standup update from current activity |
/review | Structured code review |
/debug | Structured debugging session |
/architecture | Create an Architecture Decision Record |
/incident | Start an incident-response workflow |
/deploy-checklist | Pre-deployment checklist |
Slack, Linear, Asana, Atlassian (Jira/Confluence), Notion, GitHub, PagerDuty, Datadog, Google Calendar, Gmail.
Version: 1.1.0
Accessibility Review — checks designs and code for WCAG 2.1 AA compliance.
Design Critique — assesses designs on usability, visual hierarchy, consistency, and accessibility.
Design Handoff — produces comprehensive developer-handoff documentation.
Design System Management — manages design tokens, component libraries, and pattern documentation.
User Research — plans, runs, and synthesises research.
UX Writing — writes effective microcopy.
| Command | Description |
|---|---|
/critique | Structured design feedback |
/design-system | Audit or extend the design system |
/handoff | Generate developer handoff specs |
/ux-copy | Write or review UX copy |
/accessibility | WCAG accessibility audit |
/research-synthesis | Synthesise research findings |
Version: 1.1.0
Competitive Analysis — analyses competitors with feature comparison matrices, positioning analysis, win/loss analysis.
Feature Spec — writes PRDs and feature specs with the INVEST framework.
Metrics Tracking — defines, tracks, and analyses product metrics (North Star, L1, L2).
Roadmap Management — plans and prioritises roadmaps (Now/Next/Later, RICE, MoSCoW).
Stakeholder Comms — produces stakeholder updates per audience.
User Research Synthesis — synthesises qualitative and quantitative research.
| Command | Description |
|---|---|
/write-spec | Write a feature spec or PRD |
/roadmap-update | Update or prioritise the roadmap |
/stakeholder-update | Generate a stakeholder update |
/synthesize-research | Synthesise user research |
/competitive-brief | Create a competitive analysis brief |
/metrics-review | Analyse product metrics |
Version: 1.1.0
Brand Voice — applies brand voice and checks adherence.
Campaign Planning — plans marketing campaigns with a framework (goal, audience, message, channel, measurement).
Competitive Analysis — researches competitors and compares positioning.
Content Creation — produces marketing content: blog posts, social, email newsletters, landing pages.
Performance Analytics — analyses marketing performance across channels.
| Command | Description |
|---|---|
/draft-content | Create content across channels |
/campaign-plan | Campaign planning and brief |
/brand-review | Check content against brand voice |
/competitive-brief | Competitive positioning brief |
/performance-report | Marketing performance report |
/seo-audit | Comprehensive SEO audit |
/email-sequence | Draft email sequences |
Version: 1.0.0
Brand Voice Enforcement — applies existing brand guidelines to content.
Discover Brand — autonomously searches enterprise platforms for brand materials (Notion, Confluence, Google Drive, Box, SharePoint, Slack, Gong, Granola, Figma).
Guideline Generation — generates brand guidelines from source materials.
| Command | Description |
|---|---|
/discover-brand | Search brand materials across platforms |
/generate-guidelines | Generate brand-voice guidelines from sources |
/enforce-voice | Apply brand guidelines to content |
Brand Voice requires a local config file .claude/brand-voice.local.md with company name, enabled platforms, search depth, known brand-material locations, and enforcement strictness (strict/balanced/flexible).
Plugins are enabled in the Claude Desktop app under Settings → Plugins.
After enabling, create .claude/settings.local.json:
{
"name": "Your name",
"title": "Your role",
"team": "Your team",
"company": "Company",
"tech_stack": ["Next.js", "TypeScript", "Supabase"],
"deploy_process": "GitHub Actions CI/CD"
}
Recommended connectors per plugin:
| Plugin | Must-have | Nice-to-have |
|---|---|---|
| Engineering | GitHub, Slack | Linear/Jira, PagerDuty, Datadog |
| Design | Figma, Slack | Notion |
| Product Management | Notion, Slack | Linear/Jira, Google Calendar |
| Marketing | Google Drive, Slack | Notion, Gmail |
| Brand Voice | Notion | Google Drive, Slack, Gong |
Standalone + supercharged: all commands work without connectors. Connectors make them better but aren't required.
Skills activate automatically: you don't have to invoke skills manually. Claude detects the context and activates the relevant expertise itself.
Commands are explicit: use slash commands for structured workflows.
Plugins work best together:
/write-spec → /architecture → /review → /campaign-plan → /draft-content/standup → /stakeholder-update/enforce-voice → /draft-content → /brand-review/discover-brand → /generate-guidelines → /enforce-voiceBrand Voice and Competitive Analysis outputs include confidence levels (high/medium/low). On low confidence always verify manually.
claude-plugins/
│
├── .claude/
│ ├── settings.local.json
│ ├── brand-voice-guidelines.md
│ └── brand-voice.local.md
│
├── docs/
│ ├── setup/
│ ├── plugins/
│ ├── workflows/
│ └── troubleshooting/
│
├── templates/
│ ├── prd-template.md
│ ├── adr-template.md
│ └── campaign-brief-template.md
│
├── brand/
│ ├── brand-guidelines.md
│ └── discovery-report.md
│
├── outputs/
│ ├── reviews/
│ ├── specs/
│ ├── reports/
│ ├── content/
│ └── postmortems/
│
└── connectors/
├── .mcp.json
└── CONNECTORS.md
Plugin not detected: check that the plugin is enabled under Settings → Plugins. Restart Claude.
Command returns generic output: make sure settings.local.json is configured correctly.
Connector won't connect: check the MCP config. Disconnect the connector and reconnect.
Brand Voice guidelines aren't loading: make sure .claude/brand-voice-guidelines.md exists and isn't empty.
"Missing scopes": the connector needs additional permissions. Disconnect and reconnect.
Timeout on Slack/Notion: large workspaces can cause long load times. Reduce search depth in the Brand Voice config.
Setup (required):
Plugin references (required):
Brand Voice (required for content teams):
No spam, unsubscribe at any time