mcp-server-moltmark

A Model Context Protocol (MCP) server for Moltmark agent certification. This server provides tools for managing agent trust scores, capabilities, and certification status.

Features

Agent Certification - Track and verify agent trust scores
Capability Registry - Declare and manage agent skills
Test Results - Record test outcomes that affect trust scores
Auto-Certification - Agents are automatically certified when they reach 80% trust score with 5+ tests

Installation

npm install mcp-server-moltmark

Or run directly with npx:

npx mcp-server-moltmark

Requirements

Node.js 18+
PostgreSQL database
DATABASE_URL environment variable

Configuration

Set the DATABASE_URL environment variable to your PostgreSQL connection string:

export DATABASE_URL="postgresql://user:password@localhost:5432/moltmark"

Claude Desktop Configuration

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "moltmark": {
      "command": "npx",
      "args": ["mcp-server-moltmark"],
      "env": {
        "DATABASE_URL": "postgresql://user:password@localhost:5432/moltmark"
      }
    }
  }
}

Tools

get_certification

Query the certification status of an agent, including trust score, capabilities, and test history.

get_certification({ agent_id: "agent-123" })

Returns:

Agent details (trust score, certified status)
List of declared capabilities
Recent test results
Test summary (passed/failed counts)

report_test_result

Submit a test outcome for an agent's capability. This affects the agent's trust score.

report_test_result({
  agent_id: "agent-123",
  capability: "code-generation",
  result: "pass", // or "fail"
  evidence: "Successfully generated valid TypeScript code for the given task"
})

The trust score is automatically recalculated after each test result.

declare_capability

declare_capability({
  agent_id: "agent-123",
  capability_name: "code-generation",
  description: "Ability to generate code in multiple programming languages"
})

verify_agent

Check if an agent meets a minimum trust score threshold.

verify_agent({
  agent_id: "agent-123",
  min_trust_score: 75
})

Returns whether the agent meets the threshold and the reason.

list_verified_agents

Get a list of certified agents, optionally filtered by capability.

// List all certified agents
list_verified_agents({})

// Filter by capability
list_verified_agents({ capability_filter: "code" })

Trust Score Calculation

Trust score = (passed tests / total tests) × 100
Auto-certification occurs when:
- Trust score ≥ 80%
- At least 5 tests recorded

Database Schema

The server automatically creates the following tables:

agents - Stores agent information and certification status
capabilities - Stores declared capabilities per agent
test_results - Records all test outcomes

Development

# Clone the repository
git clone https://github.com/yourusername/mcp-server-moltmark.git
cd mcp-server-moltmark

# Install dependencies
npm install

# Build
npm run build

# Run locally
DATABASE_URL="postgresql://..." npm start

License

MIT

mcp-server-moltmark