# Flow Nexus Integration Guide ## Overview Agentic Flow now integrates with **Flow Nexus**, a cloud platform for AI development, adding **96 additional MCP tools** for cloud sandboxes, distributed swarms, neural training, and more. ## Total MCP Tools: 203 | Server | Tools | Description | |--------|-------|-------------| | **claude-flow** | 101 | Core orchestration, memory, GitHub, neural networks | | **flow-nexus** | 96 | Cloud sandboxes, distributed swarms, workflows, challenges | | **claude-flow-sdk** | 6 | In-process memory and coordination tools | | **TOTAL** | **203** | Complete AI orchestration toolkit | --- ## Flow Nexus Features ### ☁️ Cloud Sandboxes (E2B) Create isolated execution environments for code development and testing: - `sandbox_create` - Create Node.js, Python, React, or Next.js sandboxes - `sandbox_execute` - Run code in isolated environment - `sandbox_upload` - Upload files to sandbox - `sandbox_status` - Check sandbox health - `sandbox_logs` - View execution logs - `sandbox_delete` - Clean up resources ### 🤖 Distributed Swarms Deploy multi-agent swarms in the cloud with auto-scaling: - `swarm_init` - Initialize cloud swarm with topology - `swarm_scale` - Scale agents up or down - `agent_spawn` - Create specialized agents - `task_orchestrate` - Distribute tasks across agents - `swarm_status` - Monitor swarm health ### 🧠 Distributed Neural Training Train neural networks across multiple cloud sandboxes: - `neural_train` - Train models with distributed compute - `neural_predict` - Run inference on trained models - `neural_cluster_init` - Initialize training cluster - `neural_node_deploy` - Deploy training nodes - `neural_training_status` - Monitor training progress ### ⚡ Workflow Automation Event-driven workflow orchestration with message queues: - `workflow_create` - Define automated workflows - `workflow_execute` - Run workflows asynchronously - `workflow_status` - Check workflow progress - `workflow_queue_status` - Monitor message queues ### 🎯 Challenges & Gamification Coding challenges with leaderboards and achievements: - `challenges_list` - Browse available challenges - `challenge_submit` - Submit solution for validation - `leaderboard_get` - View rankings - `achievements_list` - Track user achievements ### 💰 Credit Management Pay-as-you-go pricing with auto-refill: - `check_balance` - View current credit balance - `create_payment_link` - Generate payment link - `configure_auto_refill` - Set up automatic refills - `get_payment_history` - View transaction history ### 📦 Templates & App Store Pre-built project templates and marketplace: - `template_list` - Browse available templates - `template_deploy` - Deploy template to sandbox - `app_search` - Search marketplace apps - `app_get` - Get application details --- ## Authentication Flow Nexus requires authentication to access cloud features. ### Register New Account ```bash # Using CLI npx agentic-flow --agent flow-nexus-auth \ --task "Register account with email: user@example.com, password: secure123" # Or via MCP tool directly mcp__flow-nexus__user_register({ email: "user@example.com", password: "secure123", full_name: "Your Name" }) ``` ### Login to Existing Account ```bash # Using CLI npx agentic-flow --agent flow-nexus-auth \ --task "Login with email: user@example.com, password: secure123" # Or via MCP tool directly mcp__flow-nexus__user_login({ email: "user@example.com", password: "secure123" }) ``` ### Check Authentication Status ```bash npx agentic-flow --agent flow-nexus-user-tools \ --task "Check my authentication status and profile" ``` --- ## Usage Examples ### Create and Execute in Sandbox ```bash npx agentic-flow --agent flow-nexus-sandbox \ --task "Create a Node.js sandbox named 'api-dev', execute 'console.log(process.version)', and show the output" ``` ### Deploy Distributed Swarm ```bash npx agentic-flow --agent flow-nexus-swarm \ --task "Initialize a mesh topology swarm with 5 agents, then orchestrate building a REST API with authentication" ``` ### Train Neural Network ```bash npx agentic-flow --agent flow-nexus-neural \ --task "Train a classification neural network using distributed training across 3 nodes" ``` ### Create Workflow ```bash npx agentic-flow --agent flow-nexus-workflow \ --task "Create an event-driven workflow that triggers on git push, runs tests, and deploys on success" ``` ### Check Credits ```bash npx agentic-flow --agent flow-nexus-payments \ --task "Check my current credit balance and payment history" ``` --- ## Validation Results ### Local Validation ✅ ```bash node validation/test-flow-nexus.js user@example.com password123 ``` **Results:** - ✅ **203 total MCP tools discovered** - 101 from claude-flow - 96 from flow-nexus - 6 from claude-flow-sdk - ✅ **Authentication successful** - Login working - Session persistence - User profile accessible - ✅ **System health verified** - Database: Healthy - Version: 2.0.0 - All services operational - ⚠️ **Sandbox creation** (requires 10 credits minimum) ### Docker Validation ✅ ```bash docker build -t agentic-flow:flow-nexus . docker run --rm -e ANTHROPIC_API_KEY=sk-ant-... agentic-flow:flow-nexus --help ``` **Results:** - ✅ Docker image builds successfully (~5 minutes) - ✅ All three MCP servers initialized in container - ✅ CLI functional with Flow Nexus integration - ✅ 75 agents loaded including flow-nexus specialists --- ## Pricing Flow Nexus uses a credit-based system: | Resource | Cost (Credits) | |----------|----------------| | Sandbox (hourly) | 10 credits | | Swarm agent (hourly) | 5 credits | | Neural training (job) | 20-100 credits | | Workflow execution | 1-5 credits | **Credit Packages:** - Starter: $10 = 100 credits - Developer: $50 = 550 credits (10% bonus) - Professional: $200 = 2400 credits (20% bonus) **Auto-refill Available:** - Set threshold (e.g., 20 credits) - Automatic recharge when balance drops - Configurable refill amount --- ## Architecture ### Triple MCP Server Setup ```typescript // src/agents/claudeAgent.ts mcpServers: { // 1. In-SDK Server (6 tools, in-process) 'claude-flow-sdk': claudeFlowSdkServer, // 2. Claude Flow (101 tools, subprocess) 'claude-flow': { command: 'npx', args: ['claude-flow@alpha', 'mcp', 'start'] }, // 3. Flow Nexus (96 tools, subprocess) 'flow-nexus': { command: 'npx', args: ['flow-nexus@latest', 'mcp', 'start'] } } ``` ### Agent Coordination Flow Nexus agents automatically coordinate with claude-flow for: - **Memory sharing** - Store sandbox IDs and workflow results - **Swarm coordination** - Combine local and cloud agents - **Task orchestration** - Distribute work across environments - **Performance monitoring** - Track resource usage --- ## Troubleshooting ### Authentication Issues **Problem:** Login fails with "Invalid credentials" **Solution:** 1. Verify email/password are correct 2. Check if account is registered: `mcp__flow-nexus__user_register` 3. Reset password if needed: `mcp__flow-nexus__user_reset_password` ### Insufficient Credits **Problem:** "Insufficient credits to create sandbox" **Solution:** 1. Check balance: `mcp__flow-nexus__check_balance` 2. Create payment link: `mcp__flow-nexus__create_payment_link({ amount: 10 })` 3. Enable auto-refill: `mcp__flow-nexus__configure_auto_refill({ enabled: true, threshold: 20, amount: 100 })` ### MCP Server Connection **Problem:** Flow Nexus tools not appearing **Solution:** 1. Verify installation: `npm ls flow-nexus` 2. Check MCP server status in logs 3. Restart with: `npx flow-nexus@latest mcp start` ### Docker Environment **Problem:** Flow Nexus not working in Docker **Solution:** 1. Ensure `ANTHROPIC_API_KEY` is set 2. Verify network connectivity for MCP subprocesses 3. Check logs: `docker logs ` --- ## Security Notes ⚠️ **Important Security Practices:** 1. **Never hardcode credentials** in source code or Docker images 2. **Use environment variables** for sensitive data 3. **Rotate API keys regularly** for production deployments 4. **Enable 2FA** on Flow Nexus account for production use 5. **Use separate accounts** for development vs production --- ## Support - **Flow Nexus Docs**: https://github.com/ruvnet/flow-nexus - **Agentic Flow Issues**: https://github.com/ruvnet/agentic-flow/issues - **MCP Protocol**: https://modelcontextprotocol.io - **Claude Agent SDK**: https://docs.claude.com/en/api/agent-sdk --- **Ready to build with cloud-powered AI agents? 🚀**