| name | automation-orchestration |
| description | Workflow automation and multi-agent orchestration tools. Use when the user mentions n8n, Langflow, claude-squad, container-use, Dagger, workflow automation, pipeline orchestration, multi-agent coordination, or wants to set up automated workflows. Also trigger for 'automate this process', 'create a workflow', 'run agents in parallel', or integration of multiple tools into a pipeline. |
Automation & Orchestration
Tools and patterns for workflow automation, LLM pipelines, and multi-agent coordination.
n8n - Visual Workflow Automation
Self-hosted workflow automation with 400+ integrations. Connects APIs, databases, and services with a visual editor.
Setup
npx n8n
docker run -it --rm -p 5678:5678 n8nio/n8n
Core Concepts
- Nodes: Individual operations (HTTP request, transform, filter)
- Triggers: Start workflows (webhook, cron, file change, email)
- Connections: Data flow between nodes
- Credentials: Stored authentication for services
- Code Nodes: Custom JavaScript or Python for complex logic
Common Patterns
Webhook Trigger → Validate Input → Process Data → Send Result
Cron Schedule → Fetch Data → Transform → Store in DB → Notify
File Upload → Extract Text → AI Analysis → Email Report
GitHub PR → Run Tests → Post Results → Update Status
When to Use
- Connecting multiple SaaS services (Slack, GitHub, Gmail, Notion)
- Scheduled data processing and reporting
- Event-driven automation (webhook triggers)
- Non-developer-friendly workflow creation
Langflow - Visual LLM Pipeline Builder
Drag-and-drop builder for RAG pipelines, agent chains, and LLM applications.
Setup
pip install langflow
langflow run
Core Concepts
- Components: Modular building blocks (LLMs, embeddings, vector stores, tools)
- Flows: Connected component graphs forming a pipeline
- Custom Components: Python classes extending base components
- API Export: Deploy flows as REST APIs
Common Patterns
Document Loader → Text Splitter → Embeddings → Vector Store → RAG Chain
User Input → Prompt Template → LLM → Output Parser → Structured Response
Agent → Tool Selection → Execution → Memory → Response
When to Use
- Building RAG applications visually
- Prototyping LLM chains before coding them
- Non-code-first LLM pipeline development
- Quick comparison of different LLM configurations
claude-squad - Multi-Agent Terminal Orchestration
Run multiple Claude Code instances in parallel, each with its own git worktree.
Setup
go install github.com/smtg-ai/claude-squad@latest
Core Concepts
- Sessions: Independent Claude Code instances
- Worktrees: Each agent gets its own git worktree (isolated branches)
- TUI Dashboard: Monitor all agents from a single terminal
- Auto-merge: Combine agent outputs back to main branch
Usage
claude-squad
When to Use
- Large refactoring tasks that touch many independent files
- Parallel feature development across modules
- Research tasks where multiple agents investigate different angles
- Any task that decomposes into independent, parallelizable subtasks
container-use - Containerized Agent Execution (Dagger)
Run AI agents in isolated Docker containers for safe code execution.
Setup
{
"container-use": {
"command": "npx",
"args": ["-y", "@anthropic/container-use-mcp"],
"env": {
"DOCKER_HOST": "unix:///var/run/docker.sock"
}
}
}
Core Concepts
- Sandboxed execution: Agent code runs in containers, not on host
- Reproducible environments: Dockerfile defines exact dependencies
- MCP integration: Available as MCP server for Claude Code
- Resource limits: CPU, memory, network constraints per container
When to Use
- Running untrusted or experimental code safely
- Reproducible build and test environments
- Isolating agent actions from the host system
- CI/CD-like workflows triggered by AI agents
Orchestration Patterns
Fan-Out / Fan-In
Distribute work to multiple agents, aggregate results:
Coordinator
├── Agent A → Result A ─┐
├── Agent B → Result B ──┼→ Aggregate → Final Output
└── Agent C → Result C ─┘
Use when: Tasks are independent and can run in parallel.
Pipeline (Sequential Stages)
Each stage processes and passes to the next:
Research → Plan → Implement → Test → Review → Deploy
Use when: Each stage depends on the previous stage's output.
Event-Driven
Agents react to events rather than being explicitly scheduled:
File Change → Lint Agent
PR Created → Review Agent
Test Failure → Debug Agent
Deploy Success → Monitor Agent
Use when: Responses should be triggered by external events.
Supervisor / Worker
One coordinator manages many workers:
Supervisor (assigns tasks, monitors progress)
├── Worker 1 (executes assigned task)
├── Worker 2 (executes assigned task)
└── Worker 3 (executes assigned task)
Use when: Tasks need dynamic assignment and coordination.
Map-Reduce
Split input, process in parallel, combine results:
Input Data → Split into chunks
├── Process chunk 1 → Partial result
├── Process chunk 2 → Partial result
└── Process chunk 3 → Partial result
Combine partial results → Final output
Use when: Processing large datasets or documents.
Tool Comparison
| Tool | Best For | Setup | Code Required |
|---|
| n8n | SaaS integrations | Docker/npx | Low (visual) |
| Langflow | LLM pipelines | pip install | Low (visual) |
| claude-squad | Parallel Claude agents | Go binary | None |
| container-use | Safe code execution | Docker + MCP | Low |
Choosing the Right Tool
- "Connect Slack to GitHub to Notion" → n8n
- "Build a RAG pipeline with embeddings" → Langflow
- "Refactor 10 files in parallel" → claude-squad
- "Run untrusted code safely" → container-use
- "Complex multi-step agent workflow" → Combine tools as needed