Team Collaboration Features

Claude CoWork transforms AI from a solo tool into a team capability. This lesson covers every collaboration feature in detail, with practical examples showing how teams use each one effectively.

Shared Conversation Threads

Shared conversations are the foundation of team collaboration in CoWork. Unlike private conversations that only you can see, shared conversations are visible to all invited team members and persist as a team resource.

Creating a Shared Conversation

When creating a new conversation, you choose its visibility level:

Practical Example: Architecture Decision

Imagine your team is deciding between two database options. Here is how a shared conversation works in practice:

# Conversation: "Database Selection for User Service"
# Visibility: Project (Backend Team)
# Participants: Alice (Tech Lead), Bob (Senior Dev), Carol (DBA)

[Alice - 10:15 AM]
We need to choose between PostgreSQL and MongoDB for the new user
service. Key requirements:
- 50M user records
- Complex queries on user attributes
- ACID compliance for financial data
- Read-heavy workload (90/10 read/write)

Claude, analyze both options against these requirements.

[Claude]
Based on your requirements, here's my analysis...
(detailed comparison follows)

[Bob - 10:32 AM]
Good analysis. Can you also consider connection pooling implications?
We're running on Kubernetes with up to 50 pods.

[Claude]
Great point. With 50 pods, connection management becomes critical...

[Carol - 11:05 AM]
@Claude, add replication and backup strategy comparison too.
We need RPO under 1 minute.

[Claude]
Adding replication analysis with your RPO requirement...

Prompt Libraries and Templates

Prompt libraries are collections of vetted, reusable prompt templates that your team has developed and refined. They ensure consistency, save time, and encode institutional knowledge about how to get the best results from Claude.

Template Structure

Every prompt template in CoWork includes these components:

• Name: A clear, descriptive title
• Category: Engineering, Product, Design, Analytics, Operations, etc.
• Description: When and why to use this template
• Variables: Placeholders that users fill in (wrapped in {{double_braces}})
• Prompt body: The actual prompt text with variables
• Metadata: Author, version, usage count, average rating

# Bug Report Analysis
# Category: Engineering
# Author: Alice Chen | Version: 2.3 | Uses: 847
# Variables: {{bug_report}}, {{codebase_area}}

You are a senior software engineer analyzing a bug report.
Use our team's knowledge base for codebase context.

Bug Report:
{{bug_report}}

Affected Area: {{codebase_area}}

Please provide:
1. **Root Cause Analysis**: Most likely cause based on symptoms
2. **Affected Components**: Which modules/services are impacted
3. **Severity Assessment**: Critical/High/Medium/Low with reasoning
4. **Suggested Fix**: Approach to resolve, with pseudocode if helpful
5. **Regression Risk**: What could break when fixing this
6. **Test Cases**: Specific tests to verify the fix

Reference our coding standards and architecture docs for context.

Template Versioning

Templates are versioned automatically. When someone edits a template, a new version is created. You can compare versions, revert to older ones, and see a changelog of what was modified and why.

Knowledge Base Management

The team knowledge base is a persistent store of documents and information that Claude can reference during any conversation. Think of it as giving Claude long-term memory about your team, projects, and processes.

What to Include

Organizing Your Knowledge Base

Structure your knowledge base with tags and categories for efficient retrieval:

# Create categories
cowork kb category create "Architecture" --description "System design docs"
cowork kb category create "Standards" --description "Coding and process standards"
cowork kb category create "Product" --description "PRDs and feature specs"

# Upload with tags and categories
cowork kb upload --file ./docs/auth-service.md \
  --category "Architecture" \
  --tags "auth,security,backend" \
  --freshness "quarterly"  # Remind to update every quarter

# Set document priority (higher = more likely to be referenced)
cowork kb set-priority --file "coding-standards.md" --priority high

# View knowledge base health
cowork kb health
# Output:
# Total documents: 47
# Last updated: 2 days ago
# Stale documents (>90 days): 3
# Most referenced: coding-standards.md (342 references)
# Least referenced: legacy-migration-plan.md (2 references)

Real-Time Collaboration

CoWork supports real-time multi-user collaboration within conversations. Multiple team members can be active in the same conversation simultaneously, similar to collaborative document editing.

How It Works

• Presence indicators: See who is currently viewing the conversation with colored avatars
• Typing indicators: Know when a colleague is composing a message to Claude
• Sequential messaging: Messages are queued and sent to Claude in order, maintaining conversation coherence
• Branching: Fork a conversation into a separate thread to explore an idea without disrupting the main discussion

Handoff Between Team Members

Handoffs let you transfer an in-progress AI conversation to another team member with full context preserved. This is essential for tasks that span shifts, specialties, or time zones.

Handoff Workflow

1. Prepare the Handoff

   Click the Handoff button in the conversation toolbar. Add a summary note explaining the current state, what has been accomplished, and what remains.

2. Select the Recipient

   Choose a team member or a role (e.g., "next on-call engineer"). The recipient gets a notification with your handoff summary.

3. Recipient Picks Up

   The recipient opens the conversation and sees the full history plus your handoff notes, highlighted at the top. They continue where you left off.

## Handoff: API Rate Limiting Investigation
**From:** Alice (Backend Team) | **To:** Bob (Backend Team)
**Status:** In Progress

### What's Done:
- Identified rate limiting issue in the /api/v2/users endpoint
- Claude analyzed the nginx config and found the limit is set to
  100 req/min but our load tests show failures at 80 req/min
- Root cause narrowed to: connection pooling OR upstream timeout

### What's Needed:
- Check the connection pool settings in our Go service
- Test with the updated nginx config Claude suggested (see message #14)
- If the fix works, draft a PR using the code review template

### Key Context:
- Affecting 3 enterprise customers (tickets: SUPPORT-4521, 4523, 4525)
- SLA deadline: Friday EOD

Review and Approval Workflows

For sensitive outputs — like customer communications, policy documents, or production configurations — CoWork supports review and approval workflows before content is finalized.

Setting Up a Review Workflow

1. Mark a Conversation for Review

   Click Request Review and select the output that needs approval. Add context about what reviewers should check.

2. Assign Reviewers

   Select one or more reviewers. You can require all reviewers or just one to approve. Set a deadline for the review.

3. Review and Annotate

   Reviewers can approve, request changes, or add inline comments on specific parts of the AI output. Comments are visible to all reviewers.

4. Approve or Iterate

   Once approved, the output is marked as "team-approved" and can be used confidently. If changes are requested, the original author iterates with Claude and resubmits.

Comments and Annotations

Team members can add comments and annotations to any message in a shared conversation. Comments are threaded, so discussions about specific outputs stay organized and do not clutter the main conversation flow.

Types of Annotations

• Inline comments: Highlight specific text in Claude's output and add a comment (similar to Google Docs commenting)
• Reactions: Quick emoji reactions to signal agreement, flag concerns, or mark useful outputs
• Status tags: Mark outputs as "Verified," "Needs Review," "Outdated," or "Do Not Use"
• Bookmarks: Save important messages to a personal or team bookmark collection for quick reference

Version History and Audit Trails

Every action in CoWork is logged in an immutable audit trail. This is critical for compliance, debugging, and understanding how AI is being used across your organization.

What's Tracked

# View recent audit events
cowork audit list --last 24h

# Filter by user
cowork audit list --user alice@acme.com --last 7d

# Filter by event type
cowork audit list --type "template.used" --last 30d

# Export audit log for compliance
cowork audit export --from 2026-01-01 --to 2026-03-14 \
  --format csv --output audit-q1-2026.csv