Skip to main content

Model Context Protocol (MCP) Implementation in ValkyrAI

ValkyrAI implements the Model Context Protocol (MCP) as both a consumer and publisher of context-driven AI services. This protocol enables ValkyrAI's microservices, agents, and external tools to exchange critical context (e.g., code snippets, assets, user preferences) in a consistent, protocol-driven manner.

What is MCP?​

The Model Context Protocol (MCP) is a specification for packaging, transmitting, and consuming context between AI models, code generation services, user interface components, and more. MCP is designed to ensure:

  • Consistency: Standardized data structure and message format
  • Extensibility: Seamless addition of new context types (preferences, assets, code stubs, etc.)
  • Interoperability: Integration across ValkyrAI, ThorAPI, ValorIDE, and external services via a shared protocol

ValkyrAI as an MCP Consumer​

As an MCP consumer, ValkyrAI can leverage services provided by other MCP-compliant systems:

Code Generation Services​

ValkyrAI can request code generation from ThorAPI or other MCP-compliant code generators:

{
  "request": {
    "type": "code_generation",
    "language": "java",
    "framework": "spring",
    "description": "Create a REST controller for user management"
  }
}

Asset Generation​

ValkyrAI can request asset generation from MCP-compliant services:

{
  "request": {
    "type": "asset_generation",
    "asset_type": "image",
    "prompt": "A futuristic city skyline with flying vehicles",
    "parameters": {
      "width": 1024,
      "height": 768,
      "style": "photorealistic"
    }
  }
}

Data Transformation​

ValkyrAI can request data transformation services:

{
  "request": {
    "type": "data_transformation",
    "source_format": "csv",
    "target_format": "json",
    "data": "..."
  }
}

ValkyrAI as an MCP Publisher​

As an MCP publisher, ValkyrAI provides services to other MCP-compliant systems:

Workflow Management​

ValkyrAI exposes its workflow management capabilities via MCP:

{
  "service": "workflow_management",
  "operations": [
    {
      "name": "create_workflow",
      "parameters": {
        "name": "string",
        "description": "string",
        "tasks": "array"
      }
    },
    {
      "name": "start_workflow",
      "parameters": {
        "id": "string"
      }
    },
    {
      "name": "get_workflow_status",
      "parameters": {
        "id": "string"
      }
    }
  ]
}

SecureField Encryption​

ValkyrAI exposes SecureField encryption services via MCP:

{
  "service": "secure_field",
  "operations": [
    {
      "name": "encrypt",
      "parameters": {
        "data": "string",
        "field_type": "string"
      }
    },
    {
      "name": "decrypt",
      "parameters": {
        "encrypted_data": "string"
      }
    }
  ]
}

Code Generation​

ValkyrAI exposes ThorAPI code generation capabilities via MCP:

{
  "service": "code_generation",
  "operations": [
    {
      "name": "generate_spring_boot",
      "parameters": {
        "openapi_spec": "string",
        "options": "object"
      }
    },
    {
      "name": "generate_typescript_client",
      "parameters": {
        "openapi_spec": "string",
        "options": "object"
      }
    }
  ]
}

MCP Integration in ValkyrAI Components​

Workflow Engine​

The ValkyrAI Workflow Engine integrates with MCP to enable:

  • Dynamic configuration of workflow tasks based on MCP context
  • Execution of MCP service calls as part of workflow tasks
  • Publishing workflow status and results as MCP context

Example of an MCP-enabled workflow task:

{
  "id": "generate-code",
  "type": "mcp",
  "parameters": {
    "service": "code_generation",
    "operation": "generate_spring_boot",
    "parameters": {
      "openapi_spec": "${workflow.inputs.openapi_spec}",
      "options": {
        "package": "com.example",
        "generate_tests": true
      }
    }
  }
}

SecureFieldKMS​

The SecureFieldKMS integrates with MCP to enable:

  • Secure exchange of encryption keys via MCP
  • Key rotation operations triggered by MCP requests
  • Encryption and decryption services exposed via MCP

HeimdaLLM​

HeimdaLLM integrates with MCP to enable:

  • Dynamic configuration of LLM parameters via MCP
  • Sharing of LLM-generated content as MCP context
  • Orchestration of multiple LLM services via MCP

MCP Server Configuration​

ValkyrAI can be configured as an MCP server by adding the appropriate configuration to the mcp-servers.json file:

{
  "servers": [
    {
      "id": "valkyrai-workflow",
      "name": "ValkyrAI Workflow Engine",
      "description": "MCP server for ValkyrAI workflow management",
      "endpoint": "http://localhost:8080/api/mcp/workflow",
      "auth": {
        "type": "bearer",
        "token_endpoint": "http://localhost:8080/api/auth/token"
      }
    },
    {
      "id": "valkyrai-securefield",
      "name": "ValkyrAI SecureField",
      "description": "MCP server for SecureField encryption services",
      "endpoint": "http://localhost:8080/api/mcp/securefield",
      "auth": {
        "type": "bearer",
        "token_endpoint": "http://localhost:8080/api/auth/token"
      }
    }
  ]
}

MCP Client Configuration​

To configure ValkyrAI as an MCP client, add the following to the application.yaml file:

valkyrai:
  mcp:
    clients:
      - id: thorapi-codegen
        name: ThorAPI Code Generation
        endpoint: http://localhost:8081/api/mcp
        auth:
          type: bearer
          token: ${THORAPI_MCP_TOKEN}
      - id: valoride-tools
        name: ValorIDE Tools
        endpoint: http://localhost:8082/api/mcp
        auth:
          type: bearer
          token: ${VALORIDE_MCP_TOKEN}

MCP Roadmap in ValkyrAI​

ValkyrAI's MCP implementation roadmap includes:

1. ThorAPI CodeGen as an MCP service​

  • Expose ThorAPI's code generation functionality via a standard MCP endpoint
  • Enable external services to request code generation with consistent request/response structures

2. Utilize MCP as ValorIDE tooling​

  • Leverage MCP to push/pull context (e.g., code snippets, frameworks, libraries) into ValorIDE
  • Automate snippet retrieval via MCP for frictionless, standardized code injection

3. Utilize MCP as ValkyrAI tooling​

  • Enable ValkyrAI's internal modules, microservices, or agentic subroutines to communicate seamlessly
  • Allow each subcomponent to publish new context or subscribe to relevant context from other modules

4. MCP Publishing Protocol (Over RSS)​

  • Implement a simple, RSS-based method for publishing new or updated context
  • Enable easy integration with existing feed readers for external integrators

5. Component Library Generator​

  • Provide end-to-end generation of front-end or back-end component libraries as an MCP service
  • Allow users to define schemas or configs for needed components

6. Preferences Stack​

  • Create a centralized location for storing/retrieving dev or user preferences
  • Ensure consistent scaffolding across teams/projects without requiring repeated manual steps

7. Gridheim Sheetster Integration​

  • Extend Gridheim Sheetster with an MCP endpoint for formula logic, data transformations, or import/export routines
  • Enable embedding of spreadsheet functionality in any service that can speak MCP

8. Asset Generation​

  • Centralize calls to multiple ML models behind one MCP endpoint for generating images, audio, or other assets
  • Reduce complexity by providing a single interface that routes to the correct model

9. Dynamic Inference Stack Configuration​

  • Enable agents to reconfigure or switch inference models on-the-fly through MCP requests
  • Adapt to changing tasks in real time by switching between different specialized models

Best Practices for MCP in ValkyrAI​

Security Considerations​

  • Always use authentication for MCP endpoints
  • Encrypt sensitive data in MCP requests and responses
  • Implement rate limiting to prevent abuse
  • Validate all incoming MCP requests

Performance Optimization​

  • Use caching for frequently requested MCP context
  • Implement pagination for large context responses
  • Consider asynchronous processing for time-consuming operations
  • Monitor MCP endpoint performance

Integration Guidelines​

  • Document all MCP endpoints and operations
  • Provide example requests and responses
  • Implement versioning for MCP endpoints
  • Include error handling and detailed error messages

Conclusion​

The Model Context Protocol (MCP) is a powerful feature of ValkyrAI that enables seamless integration with other AI-powered services and tools. By implementing MCP as both a consumer and publisher, ValkyrAI can leverage external services and provide its own capabilities to other systems, creating a rich ecosystem of interoperable AI tools.

For more information on MCP, refer to the Model Context Protocol specification and the ValkyrAI API documentation.