Llama Maverick Hub MCP

You can run a centralized MCP hub that uses Llama Maverick as the AI brain to orchestrate and route requests to multiple MCP-enabled services like Stripe, GitHub, and databases. This hub exposes a unified interface to MCP clients while coordinating real services through AI-driven workflows, routing, and error handling.

How to use

When you connect an MCP client such as Claude Desktop, you gain access to tools from all connected services through a single hub. Start by ensuring the hub is running and that your external MCP services are configured. You can then compose multi-service workflows, route requests intelligently, and execute parallel queries to gather data from Stripe, GitHub, and databases. The hub will coordinate steps, pass context between stages, handle retries, and synthesize results into a single response for the client.

How to install

# Prerequisites
node -v
npm -v
# Ensure Node.js 18+ is installed

# Ollama and Llama model setup (required for local Llama integration)
curl -fsSL https://ollama.ai/install.sh | sh
ollama pull llama3.2

# Install MCP hub (from source flow shown in configuration)
git clone https://github.com/yobieben/llama-maverick-hub-mcp.git
cd llama-maverick-hub-mcp
npm install
npm run build
cp config.example.json config.json
npm start

Additional setup and usage notes

Configure the hub to connect to your desired MCP services. The hub supports multiple services via standard MCP protocol channels. Each service entry defines how the hub starts and reconnects if a connection is interrupted. You can adjust orchestration settings to control concurrency, timeouts, and retry behavior.

Configuration and security

Set environment variables and authentication as needed to protect access and provide tokens to connected services. Example variables include tokens for GitHub and Stripe integrations. Always isolate services in separate processes and enable audit logging for workflow executions.

Troubleshooting and tips

If a service fails to connect, verify the command and arguments in the service configuration, confirm the service is enabled, and check that any required environment variables are provided. For slow or failing workflows, review provenance and dependencies between steps, and consider raising timeouts or adjusting retry policies.

Real-world patterns

Create a complete customer lifecycle workflow that touches Stripe for payments, a database for records, and email for onboarding, all orchestrated by Llama Maverick. Use parallel queries to fetch data from multiple sources and then synthesize a final report for the client.

Available tools

hub_execute_workflow

Execute a predefined multi-service workflow by coordinating steps across connected MCP services.

hub_smart_route

Use Llama AI to intelligently route a request to the most appropriate MCP service based on context.

hub_parallel_query

Query multiple services in parallel to gather data quickly and efficiently.

stripe_create_customer

Create a new Stripe customer within a managed workflow.

stripe_create_charge

Create a charge or payment intent within a workflow.

github_create_repo

Create a new repository via the GitHub MCP service as part of a workflow.

github_create_issue

Open or manage issues in a GitHub repository through an AI-guided flow.

database_query

Perform a data query against the connected database MCP.

database_insert

Insert records into the database MCP as part of a workflow.