Skip to content

ATaylorAerospace/MCP-Agent-Accelerator

πŸ€– MCP-Agent-Accelerator πŸš€

License Stars Language Contact A Taylor

A production ready framework for building AI agents using the Model Context Protocol (MCP), paired with a full-featured Next.js 14 intelligence dashboard. The accelerator eliminates integration boilerplate, provides robust agent orchestration, and ships with a working UI out of the box.

🚧 Status: Core framework stable · Dashboard live


πŸ€” The Problem

Building AI agents is more than just connecting to an LLM. Developers face real hurdles when making agents interact with production tools:

  • πŸ’₯ Complex Connections: Every new tool requires a custom, brittle integration.
  • πŸŒ€ Coordination Chaos: Making multiple agents collaborate is a major challenge.
  • ⚠️ Production Pitfalls: Moving from demo to reliable, production-grade system is hard.
  • πŸ”₯ Error Handling Hell: Managing failures and retries across services is a nightmare.

πŸ’‘ The Solution

MCP-Agent-Accelerator provides the infrastructure layer that handles complex integrations, allowing focus on unique agent logic and business value. The included Agent Accelerator Dashboard delivers a working AI interface immediately β€” no UI build required.


πŸ“ Repository Layout

MCP-Agent-Accelerator/
β”œβ”€β”€ src/                        # Core TypeScript/Python framework
β”‚   β”œβ”€β”€ main.ts                 # AgentAccelerator class (TypeScript)
β”‚   β”œβ”€β”€ main.py                 # AgentAccelerator class (Python)
β”‚   └── index.test.ts           # Jest unit tests
β”œβ”€β”€ nexus-dashboard/            # Agent Accelerator Dashboard (Next.js 14)
β”‚   └── src/
β”‚       β”œβ”€β”€ app/                # App Router pages + API routes
β”‚       β”œβ”€β”€ components/         # Shared UI components
β”‚       └── lib/                # Anthropic SDK client
β”œβ”€β”€ Dockerfile                  # Multi-stage build for the core framework
β”œβ”€β”€ Dockerfile.dashboard        # Standalone build for the Next.js dashboard
└── package.json                # Root workspace scripts

πŸ–₯️ Agent Accelerator Dashboard

The nexus-dashboard/ directory contains a production-ready Next.js 14 web application with three intelligence modules:

Module Status Description
πŸ” Web Search βœ… Live Real-time web search powered by Anthropic's web_search tool
πŸ‘₯ CRM ⚠️ Needs Setup Salesforce contact list + Kanban pipeline β€” MCP stub ready to activate
πŸ—„οΈ Data Warehouse ⚠️ Needs Setup Natural-language queries over Snowflake/BigQuery β€” MCP stub ready to activate

⚑ Dashboard quick start:

cd nexus-dashboard
npm install
cp .env.example .env.local
# Set ANTHROPIC_API_KEY in .env.local
npm run dev
# Open http://localhost:3000

Or from the root with the workspace script:

npm run dashboard:install
npm run dashboard

πŸ”— Activating CRM (Salesforce MCP)

In nexus-dashboard/.env.local:

SALESFORCE_MCP_URL=https://mcp.salesforce.com/sse
SALESFORCE_CONNECTED=true

Restart the server. The CRM page will route live prompts through the MCP server.

πŸ—„οΈ Activating Data Warehouse (Snowflake / BigQuery MCP)

In nexus-dashboard/.env.local:

DW_MCP_URL=https://your-dw-mcp-url/sse
DW_CONNECTED=true

Restart the server. Natural-language queries will be sent to the MCP server.


✨ Core Framework Capabilities

  • πŸ”§ Universal Tool Integration: Connect to anything with an MCP server. Pre-configured connections for GitHub, databases, and local file systems.
  • 🀝 Advanced Multi-Agent Orchestration: Sophisticated workflows with teams of specialized agents that collaborate to solve complex problems.
  • ⚑ Production-Ready from Day One: Comprehensive logging (Winston), monitoring hooks, and full type safety throughout.
  • πŸ“¦ Containerized and Scalable: Docker multi-stage builds for both the framework and the dashboard.

▢️ Framework Example Usage

import AgentAccelerator, { ServerConfig, WorkflowOptions } from './src/main';

const servers: ServerConfig[] = [
  { name: "github" },
  { name: "sonarqube", url: "http://localhost:9000" },
  { name: "eslint" },
  { name: "jest" }
];

const accelerator = new AgentAccelerator();
accelerator.registerServers(servers);

const options: WorkflowOptions = { timeout: 30000, retries: 3 };
accelerator.runWorkflow("code-review-workflow", options);

🏁 Getting Started

Prerequisites

  • βœ… Node.js 18+
  • βœ… npm 8+
  • βœ… Anthropic API Key (for the dashboard and Anthropic-powered workflows)
  • βœ… GitHub Token (for GitHub MCP integration)
  • βœ… Docker (optional)

πŸ› οΈ Framework Setup

# Clone the repository
git clone https://github.com/ATaylorAerospace/MCP-Agent-Accelerator.git
cd MCP-Agent-Accelerator

# Install framework dependencies
npm install

# Type-check
npm run typecheck

# Build
npm run build

# Run tests
npm test

# Run tests with coverage
npm run test:coverage

πŸ–₯️ Dashboard Setup

# Install dashboard dependencies
npm run dashboard:install

# Start the dashboard dev server
npm run dashboard

# Build the dashboard for production
npm run dashboard:build

πŸ› οΈ Technology Stack

Layer Technology
πŸ”€ Framework languages TypeScript, Python
πŸ–₯️ Dashboard Next.js 14 (App Router), React 18, Tailwind CSS
πŸ€– AI provider Anthropic (claude-sonnet-4-20250514)
πŸ“‘ Protocols MCP, gRPC, REST
🎨 Icons Lucide React
πŸ”€ Fonts DM Sans, JetBrains Mono
πŸ“‹ Logging Winston (structured JSON)
πŸ§ͺ Testing Jest + ts-jest
🐳 Infrastructure Docker, Kubernetes

πŸ›οΈ Architecture

The project separates concerns across two distinct layers:

βš™οΈ Framework core (src/) β€” Protocol-agnostic agent orchestration. The AgentAccelerator class handles MCP server registration, workflow routing, and structured logging. Python and TypeScript implementations share the same interface contract.

πŸ–₯️ Dashboard (nexus-dashboard/) β€” A Next.js 14 App Router application that consumes the Anthropic SDK directly from API routes and connects to MCP-enabled services (Salesforce, Snowflake/BigQuery) via environment-gated server stubs. The UI uses a dark design system built on CSS custom properties, Tailwind CSS, and a shared component library.


πŸ™ Contributing

Contributions of all kinds are welcome. See CONTRIBUTING.md for details.

  • πŸ”Œ New Connectors: Help integrate more tools via MCP.
  • 🀝 Orchestration Patterns: Share workflow ideas.
  • πŸ–₯️ Dashboard Modules: Add new intelligence panels.
  • πŸ“– Examples and Docs: Improve developer experience.
  • πŸš€ Performance Tuning: Make the framework faster and more efficient.

πŸ‘€ Author

A Taylor

Contact A Taylor

πŸ“„ License

MIT Β© A Taylor

About

A flexible framework that helps AI agents to connect seamlessly with real world data and tools using the Model Context Protocol

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors