Agentic DesignPatterns and tools for production AI agents

Loading...

📝

Shared Scratchpad Collaboration(SSC)

Multiple agents collaborate transparently on a common workspace with shared visibility of all work

Complexity: mediumMulti-Agent

In 30 seconds

What: Multiple agents write to and read from a shared workspace, each seeing all prior work and building incrementally on it.
When to use: Complex documents or outputs requiring different expertise, where agents benefit from seeing each other's contributions before adding their own.
Watch out: Agents may contradict or overwrite each other's work if conflict resolution and ordering rules are not explicit.

Loading technique guide…

Patterns

Loading...

Loading Agentic Design…

Agentic DesignPatterns and tools for production AI agents

Expert help Pricing

Design Patterns & Techniques

Loading...

📝

Shared Scratchpad Collaboration(SSC)

Multiple agents collaborate transparently on a common workspace with shared visibility of all work

Complexity: mediumMulti-Agent

In 30 seconds

What: Multiple agents write to and read from a shared workspace, each seeing all prior work and building incrementally on it.
When to use: Complex documents or outputs requiring different expertise, where agents benefit from seeing each other's contributions before adding their own.
Watch out: Agents may contradict or overwrite each other's work if conflict resolution and ordering rules are not explicit.

30-Second Overview

Pattern: Multiple agents collaborate transparently on a common workspace with shared visibility of all work

Why: Eliminates information silos, enables natural collaboration flow, and provides transparent decision-making through real-time visibility

Key Insight: Shared state + real-time visibility + progressive building + transparent process = natural collaborative intelligence

Quick Implementation

1Initialize:Create shared state/workspace accessible to all agents

2Define:Establish collaboration protocols and update formats

3Coordinate:Agents read, contribute, and build upon shared work

4Synchronize:Real-time visibility of all agent contributions

5Converge:Collaborative refinement toward final outcome

Example: shared_workspace.init() → agents.collaborate() → real_time_updates() → iterative_refinement()

Do's & Don'ts

✅Design clear shared state schema with structured sections

✅Implement real-time visibility of all agent contributions

✅Use version control and conflict resolution mechanisms

✅Establish clear protocols for agent contribution formats

✅Enable progressive building and iterative refinement

✅Implement proper access control and edit permissions

✅Maintain clear audit trails of all contributions

✅Design for transparent decision-making processes

❌Allow unstructured or conflicting update formats

❌Create information silos between agents

❌Skip conflict resolution for simultaneous edits

❌Ignore the computational overhead of shared state

❌Allow agents to overwrite each other without coordination

❌Forget to implement rollback mechanisms for errors

When to Use

Use When

• Collaborative content creation workflows
• Transparent decision-making processes
• Multi-perspective problem solving
• Iterative refinement requirements
• Cross-agent learning scenarios
• Consensus-building applications

Avoid When

• Independent parallel processing needs
• High-security compartmentalized tasks
• Simple linear workflow requirements
• Resource-constrained environments
• Real-time performance critical systems
• Single-agent sufficient scenarios

Key Metrics

Collaboration Efficiency

% time agents building vs duplicating work

Transparency Score

Visibility of decision-making process

Convergence Rate

Time to reach collaborative consensus

Cross-Agent Learning

% insights shared between agents

State Synchronization

Real-time update propagation speed

Conflict Resolution

% edit conflicts resolved automatically

Progressive Quality

Output improvement over iterations

Top Use Cases

Collaborative Document Creation: Multiple writing agents (research, content, editing, review) working on shared document with real-time visibility

Software Architecture Design: Architecture, security, performance, and UX agents collaboratively designing system with shared blueprint

Strategic Planning: Market, financial, operational, and risk agents building comprehensive business strategy in shared workspace

Scientific Research: Literature, methodology, analysis, and peer review agents collaborating on research paper with transparent process

Product Development: User research, design, engineering, and QA agents iteratively building product requirements in shared specification

Legal Document Review: Subject matter, compliance, risk, and editorial agents collaborating on contract analysis with full visibility

Creative Brainstorming: Concept, narrative, visual, and technical agents building creative project with transparent ideation process

Educational Curriculum: Subject expert, pedagogy, assessment, and accessibility agents designing course with collaborative refinement

References & Further Reading

Deepen your understanding with these curated resources

Academic Research

LangGraph Multi-Agent Workflows and Shared State (2024)

Implementation Frameworks

LangGraph Shared State Documentation

CrewAI Collaborative Process Patterns

AutoGen Selector Group Chat

Google ADK Agent State Management

Tools & Libraries

LangGraph - Graph-based Multi-Agent Orchestration

CrewAI - Agent Collaboration Framework

AutoGen - Microsoft Multi-Agent Framework

Swarm - OpenAI Lightweight Multi-Agent Framework

Design Patterns & Best Practices

Shared State Design Patterns for AI Systems

Conflict Resolution in Multi-Agent Systems

Real-time Collaboration Patterns for AI Agents

Community & Research

Multi-Agent Systems Research Community

LangChain Multi-Agent Community

Collaborative AI Working Group

AI Coordination Patterns Forum

Loading technique guide…

Patterns

Design Patterns & Techniques

Loading...

30-Second Overview

Pattern: Multiple agents collaborate transparently on a common workspace with shared visibility of all work

Why: Eliminates information silos, enables natural collaboration flow, and provides transparent decision-making through real-time visibility

Key Insight: Shared state + real-time visibility + progressive building + transparent process = natural collaborative intelligence

Quick Implementation

1Initialize:Create shared state/workspace accessible to all agents

2Define:Establish collaboration protocols and update formats

3Coordinate:Agents read, contribute, and build upon shared work

4Synchronize:Real-time visibility of all agent contributions

5Converge:Collaborative refinement toward final outcome

Example: shared_workspace.init() → agents.collaborate() → real_time_updates() → iterative_refinement()

Do's & Don'ts

✅Design clear shared state schema with structured sections

✅Implement real-time visibility of all agent contributions

✅Use version control and conflict resolution mechanisms

✅Establish clear protocols for agent contribution formats

✅Enable progressive building and iterative refinement

✅Implement proper access control and edit permissions

✅Maintain clear audit trails of all contributions

✅Design for transparent decision-making processes

❌Allow unstructured or conflicting update formats

❌Create information silos between agents

❌Skip conflict resolution for simultaneous edits

❌Ignore the computational overhead of shared state

❌Allow agents to overwrite each other without coordination

❌Forget to implement rollback mechanisms for errors

When to Use

Use When

• Collaborative content creation workflows
• Transparent decision-making processes
• Multi-perspective problem solving
• Iterative refinement requirements
• Cross-agent learning scenarios
• Consensus-building applications

Avoid When

• Independent parallel processing needs
• High-security compartmentalized tasks
• Simple linear workflow requirements
• Resource-constrained environments
• Real-time performance critical systems
• Single-agent sufficient scenarios