Frequently Asked Questions

MeshAI is a universal orchestration platform for AI agents. It acts as an intelligent routing layer that will connect AI agents built on different frameworks (LangChain, CrewAI, AutoGen, OpenAI, Anthropic, etc.) into one unified network.

Think of it as a smart dispatcher that automatically finds the best agent for any task, regardless of the underlying technology. We're building the infrastructure to make AI agents work together seamlessly.

MeshAI offers several key advantages:

• Framework Independence - Use any AI agent without worrying about compatibility
• Intelligent Routing - Tasks automatically go to the most capable agent
• No Vendor Lock-in - Switch between providers seamlessly
• Load Balancing - Distribute work across multiple agents efficiently
• Unified API - One interface for all your AI agents
• Built-in Monitoring - Track performance and costs in real-time

MeshAI is perfect for developers building AI applications, enterprises managing multiple AI tools, startups that need to scale quickly without rewriting code, and anyone frustrated with framework compatibility issues. If you're using multiple AI agents or planning to, MeshAI will save you significant time and complexity.

MeshAI uses intelligent capability matching. When you submit a task, our system:

1. Analyzes the task to detect required capabilities (text generation, data analysis, code generation, etc.)
2. Identifies available agents with matching capabilities
3. Evaluates agent performance metrics and current load
4. Routes to the best available agent based on your routing strategy (performance, cost, or round-robin)
5. Handles retries and fallbacks automatically if needed

Task routing in MeshAI follows a sophisticated process:

1. Capability Detection: Identifies what skills are needed (e.g., 'analyze data' triggers data_analysis capability)

2. Agent Discovery: The routing engine queries the Agent Registry for matching agents

3. Evaluation: Considers factors like agent health, recent performance, current load, and cost

4. Routing Strategies:

• performance - Fastest response time
• cost_optimized - Lowest price
• capability_match - Best skill alignment
• sticky_session - Maintain context with same agent

MeshAI supports unlimited concurrent agents. The platform is designed to scale horizontally, meaning you can run as many agents as your infrastructure supports. Our load balancing automatically distributes tasks across available agents, and our circuit breaker pattern prevents overloading. Enterprise plans include dedicated resources for guaranteed performance at scale.

MeshAI has built-in resilience features to ensure your tasks complete successfully:

• Automatic Retry - With exponential backoff for transient failures
• Intelligent Fallback - Routes to alternative agents with similar capabilities
• Circuit Breakers - Isolates failing agents to prevent cascade failures
• Dead Letter Queues - Captures failed tasks for manual review
• Real-time Monitoring - Detects issues early with health checks

Your tasks complete successfully even when individual agents experience problems.

The Prompt Optimizer is an AI-powered service that automatically improves your prompts for better performance. It uses advanced techniques to optimize prompts:

How it works:

1. Genetic Algorithms - Evolves prompt variations through mutation and selection
2. A/B Testing - Tests different prompt versions against real tasks
3. Performance Analysis - Measures accuracy, response time, and cost
4. Continuous Learning - Improves prompts based on usage patterns

Benefits:

• 15-25% performance improvement on average
• Reduced token usage and cost
• Better consistency in responses
• Automatic adaptation to different tasks

The optimizer works with any prompt format and learns from your specific use cases.

Topology Search is an advanced feature that discovers optimal multi-agent workflows for complex tasks. It finds the best combination and sequence of agents to solve your problems efficiently.

Key capabilities:

• MCTS Search - Monte Carlo Tree Search for exploring agent combinations
• Topology Discovery - Automatically maps optimal agent relationships
• Swarm Intelligence - Uses collective behavior patterns to find solutions
• Performance Validation - Tests discovered topologies before deployment

Use cases:

• Complex data analysis requiring multiple processing steps
• Content creation that needs both generation and refinement
• Research tasks requiring information gathering and synthesis
• Customer service workflows with escalation patterns

Topology Search is most valuable for tasks that would benefit from multiple specialized agents working together.

The Prompt Optimizer integrates seamlessly with your existing MeshAI workflows:

Integration methods:

• Automatic Optimization - Enable optimization for specific agent types
• Manual API Calls - Send prompts directly to the optimizer
• Batch Processing - Optimize multiple prompts at once
• Real-time Learning - Optimizer learns from successful patterns

Setup example:// Enable optimization for an agent { "agent_id": "my-agent", "enable_optimization": true, "optimization_target": "accuracy" }

The optimizer works in the background, improving prompts without requiring changes to your existing code.

Yes! Topology Search offers multiple customization options to suit your specific needs:

Customization options:

• Search Algorithms - Choose between MCTS, genetic algorithms, or hybrid approaches
• Optimization Goals - Prioritize speed, cost, accuracy, or custom metrics
• Agent Constraints - Limit which agents can be used in topologies
• Topologies - Define preferred patterns (sequential, parallel, hierarchical)
• Performance Thresholds - Set minimum performance requirements

Example configuration:{ "search_algorithm": "mcts", "optimization_target": "cost_efficiency", "max_topology_depth": 5, "allowed_agents": ["data-analyst", "summarizer", "validator"], "performance_threshold": 0.85 }

Advanced users can even implement custom search algorithms using our SDK.

This error occurs when trying to use OpenAI's MCP server alongside MeshAI's MCP server. The OAuth authentication required by OpenAI conflicts with other MCP connections.

Solution: Use only MeshAI's MCP server in your configuration. MeshAI will handle routing to OpenAI and other providers internally:{ "mcpServers": { "meshai": { "command": "npx", "args": ["-y", "meshai-mcp-server"], "env": { "MESHAI_API_KEY": "YOUR_API_KEY" } } // Don't add OpenAI server here } }
📚 See full troubleshooting guide

MCP (Model Context Protocol) integration allows you to use MeshAI through tools in your AI applications.

Configuration example:{ "mcpServers": { "meshai": { "command": "npx", "args": ["-y", "meshai-mcp-server"], "env": { "MESHAI_API_KEY": "YOUR_API_KEY" } } } }
Once configured, you can submit tasks using the mesh_execute tool, which automatically handles capability detection and routing.

📚 For detailed setup instructions, see our MCP Capabilities Guide

MeshAI is building support for all major AI frameworks. Our roadmap includes:

• LangChain
• CrewAI
• AutoGen
• OpenAI GPT
• Anthropic Claude
• Google Gemini
• And many more...

We're actively expanding our framework support. Our generic REST adapter allows you to connect any HTTP-based agent while we build native integrations.

In most cases, no modifications are needed. MeshAI adapters handle the translation between your agent's native format and our unified protocol. If you're building new agents, our SDKs provide native integration. For existing agents, you just need to register them with their endpoint URL and capabilities.

Absolutely! MeshAI is designed to work with any agent that can receive and respond to requests. You can register custom agents by providing an HTTP endpoint and defining their capabilities. Our platform handles all the routing and orchestration, regardless of how your agent is implemented internally.

MeshAI offers flexible pricing tiers: Free tier for developers to get started, Pay-as-you-go for growing applications, and Enterprise plans with dedicated support and SLAs. Pricing is based on the number of task executions, not the number of agents. This means you can connect unlimited agents and only pay for what you use.

Yes! Our free tier includes up to 1,000 task executions per month, access to all core features, community support, and basic monitoring. It's perfect for testing, development, and small projects. No credit card required to get started.

The MeshAI dashboard provides comprehensive insights:

• Real-time Tracking - Monitor task execution as it happens
• Agent Metrics - Health status, performance, and success rates
• Cost Analysis - Usage trends and spending breakdowns
• Error Logs - Detailed debugging tools and stack traces
• API Management - Create, rotate, and manage API keys
• Capability Reports - See which capabilities are most used
• Custom Alerts - Get notified about important events

Enterprise plans include advanced analytics and custom dashboards.