DANP-Engine: Trusted AI MCP Runtime

DANP-Engine is a trusted runtime for AI Model Context Protocol (MCP), providing a secure execution environment for decentralized AI tools and services. Built on four foundational technologies:

• IPFS: Decentralized storage for immutable WASM MCP AI tools modules
• WASM: Portable, sandboxed execution of AI workloads
• AI MCP Server: Hosts and manages registered AI tools
• AI MCP Client: Provides standardized access to AI capabilities

As an AI MCP runtime, DANP-Engine enables:

• Trusted execution of AI tools via WASM sandboxing
• IPFS Decentralized Verifiable Storage WASM Tool for AI MCP Server
• Standardized interfaces via MCP protocol

System Architecture

graph TD
    %% DANP-Engine Architecture Diagram
    DANP[DANP-Engine\nTrusted AI MCP Runtime] --> IPFS
    DANP --> WASM
    DANP --> MCPServer[AI MCP Server]
    DANP --> MCPClient[AI MCP Client]

    %% IPFS Component
    subgraph IPFS[IPFS Storage]
        direction LR
        ipfs_node1[IPFS Node]
        ipfs_node2[IPFS Node]
        ipfs_node3[IPFS Node]
        ipfs_node1 <--> ipfs_node2
        ipfs_node2 <--> ipfs_node3
    end

    %% WASM Component
    subgraph WASM[WASM Runtime]
        direction LR
        wasm1[WASM Module]
        wasm2[WASM Module]
        wasm3[WASM Module]
    end

    %% MCP Server Component
    subgraph MCPServer[AI MCP Server]
        direction LR
        server1[Server Node]
        server2[Server Node]
        server1 <--> server2
    end

    %% MCP Client Component
    subgraph MCPClient[AI MCP Client]
        direction LR
        client1[Client]
        client2[Client]
        client3[Client]
    end

    %% Connections
    IPFS --> WASM
    WASM --> MCPServer
    MCPServer --> MCPClient

    %% Styling
    style DANP fill:#ffebee,stroke:#333,stroke-width:2px
    style IPFS fill:#e3f9ff,stroke:#333
    style WASM fill:#fff2e6,stroke:#333
    style MCPServer fill:#e6ffe6,stroke:#333
    style MCPClient fill:#f9e6ff,stroke:#333

Core Components and Features

IPFS Integration Layer

• Decentralized Storage: All WASM modules and AI tools are stored on IPFS with content addressing
• Immutable Artifacts: Ensures tool integrity via cryptographic hashes
• Global Distribution: Tools are available from any IPFS node worldwide

WASM Runtime Layer

• Secure Sandboxing: Isolates tool execution for safety
• Cross-platform: Runs anywhere WASM is supported
• High Performance: Near-native execution speed

AI MCP Server

• Tool Hosting: Manages lifecycle of registered AI tools
• Discovery Service: Enables tool lookup and metadata access
• Execution Engine: Runs WASM modules with resource controls

AI MCP Client

• Standard Interface: Uniform access to all registered tools
• Session Management: Handles authentication and state
• Multi-client Support: CLI, Web, and programmatic access

How It's Made

DANP-Engine is built on four core components that work together to provide a trusted AI MCP runtime:

IPFS Integration

• Role: Provides decentralized, immutable storage for WASM modules and AI tools
• Implementation:
  • Integrated IPFS nodes for distributed content addressing
  • Uses Filecoin-Lassie for efficient IPFS file retrieval
  • Supports IPFS Car file extraction via Filecoin-IPLD-Go-Car

WASM Runtime

• Role: Executes trusted, portable code in a secure sandbox
• Implementation:
  • Leverages wazero for efficient WASM execution
  • Uses Extism for WASM plugin management
  • Supports both local and IPFS-hosted WASM modules

AI MCP Server

• Role: Hosts and manages AI tools and services
• Implementation:
  • Built with Fiber for high-performance HTTP serving
  • Provides tool registration and discovery
  • Manages WASM module lifecycle and execution

AI MCP Client

• Role: Interfaces with the MCP Server and provides user access
• Implementation:
  • Supports multiple client implementations (CLI, Web, etc.)
  • Provides tool discovery and invocation
  • Handles authentication and session management

Integrated Benefits

• Trusted Execution: Combines IPFS immutability with WASM sandboxing
• Decentralized AI: Enables distributed AI tool hosting and execution
• Interoperability: Standard MCP protocol connects all components

Quick Start

1. Clone the Repository

git clone https://github.com/DANP-LABS/DANP-Engine.git
cd DANP-Engine

2. Configure Your Wallet

The MCP Server requires a Web3 wallet to operate. The wallet's private key is encrypted in a wallet.json file, which is protected by a password.

Set the Wallet Password

You must provide this password to the server via the WALLET_PASSWORD environment variable.

You can set it directly in your terminal:

export WALLET_PASSWORD="your-strong-password-here"

Alternatively, you can create a .env file in the cmd/DANP-MCP-SERVER/ directory. This file is ignored by Git, so your password will not be committed.

Example cmd/DANP-MCP-SERVER/.env file:

# This password encrypts your wallet.json file.
# Keep this password and your wallet file secure.
WALLET_PASSWORD="your-strong-password-here"

When the server starts for the first time, it will automatically generate a config/wallet.json file for you using this password. Do not commit config/wallet.json to version control.

3. Build the Project

The project provides several Makefile targets for building and development:

Basic Builds

# Build both client and server
make all

# Build just the client
make build-client

# Build just the server 
make build-server

Cross-Compilation

# Build for all platforms (Linux, Windows, macOS, ARM)
make build-all

# Platform-specific builds
make build-linux    # Linux amd64
make build-windows  # Windows amd64 (.exe)
make build-darwin   # macOS amd64  
make build-arm      # Linux ARM64

Development

# Run client directly (no build)
make run-client

# Run server directly (no build)
make run-server

# Clean build artifacts
make clean

Build flags include version information:

• BuildVersion: Short git commit hash
• BuildDate: UTC timestamp of build

3. Adjust Configuration File

# MCP Server Manifest
server_config:
  host: "0.0.0.0"
  port: 18080
  max_connections: 100
  timeout: 30s

ipfs:
  enable: true  # Set to true to enable IPFS support
  lassie_net:
    scheme: "http"  # http or https
    host: "127.0.0.1"
    port: 31999
  cids: []  # Optional list of pre-loaded CIDs

llm_config:
  base_url: ""  # Optional base URL for API endpoints
  provider: "openai"  # Default provider
  openai:  # OpenAI-specific config
    api_key: ""
    model: "gpt-4"
    temperature: 0.7
    max_tokens: 2048
  # Add other provider configs here as needed

# Defines WASM modules and their exposed MCP tools
modules:
  - name: "hello"
    #wasm_path: "file://config/hello.wasm"  # Supports file:// or IPFS:// schemes
    wasm_path: "IPFS://QmeDsaLTc8dAfPrQ5duC4j5KqPdGbcinEo5htDqSgU8u8Z"  # Supports file:// or IPFS:// schemes
    tools:
      - name: "say_hello"
        description: "Greet someone by name"
        inputs:
          - name: "name"
            type: "string"
            required: true
            description: "Name to greet"
        outputs:
          type: "string"
          description: "Greeting message"

4. Load Configuration and Run MCP Server

go run cmd/DANP-MCP-SERVER/main.go

5. Interact with MCP Server using Client

go run cmd/DANP-MCP-CLIENT/main.go -http http://localhost:18080/

6. Example AI Interaction

# Server startup log showing WASM module loading from IPFS
2025/06/29 14:06:10 Loading WASM module from IPFS CID: QmeDsaLTc8dAfPrQ5duC4j5KqPdGbcinEo5htDqSgU8u8Z
2025/06/29 14:06:10 Successfully loaded WASM module: IPFS://QmeDsaLTc8dAfPrQ5duC4j5KqPdGbcinEo5htDqSgU8u8Z
2025/06/29 14:06:10 Registering tool: say_hello
2025/06/29 14:06:10 MCP server listening on 0.0.0.0:18080

# Client interaction example
Enter your request (empty line to submit, 'exit' to quit):
> Could you please greet my friend John for me?
> 

AI Response:
I've greeted your friend John for you! Here's the message: 

👋 Hello John

Contributing

We welcome contributions from the community! To contribute to DANP-Engine:

1. Fork the repository.
2. Create a new branch (git checkout -b feature/your-feature).
3. Make your changes and commit them (git commit -am 'Add new feature').
4. Push your changes to the branch (git push origin feature/your-feature).
5. Create a new Pull Request.