A Model Context Protocol (MCP) server that enables AI agents to query trust scores, attestations, and reputation data from Intuition's decentralized knowledge graph.

This project provides a standardized interface for AI agents like Claude and ChatGPT to access Intuition's attestation system. Built using the Model Context Protocol, it exposes four core tools for querying trust data and verifying credentials on-chain.

• Trust Score Calculation - Aggregate trust scores for Ethereum addresses based on attestation patterns
• Attestation Queries - Search and filter attestations with flexible parameters
• Credential Verification - Verify specific claims and credentials for addresses
• Expert Discovery - Find highly trusted addresses in specific domains
• ENS Support - Resolve ENS names to Ethereum addresses
• RESTful API - HTTP endpoints for direct integration
• Interactive Dashboard - Web-based testing interface

• Next.js 14 (App Router)
• TypeScript
• MCP SDK (@modelcontextprotocol/sdk)
• GraphQL (graphql-request)
• Tailwind CSS + shadcn/ui
• Vercel (deployment)

git clone https://github.com/rudazy/Intuition-.git
cd Intuition-
npm install

Create a .env.local file:

NEXT_PUBLIC_INTUITION_GRAPH_URL=https://mainnet.intuition.sh/v1/graphql
NEXT_PUBLIC_CHAIN_ID=1155
NEXT_PUBLIC_RPC_URL=https://rpc.intuition.systems/http
NEXT_PUBLIC_APP_URL=http://localhost:3000

npm run dev

Visit http://localhost:3000 to access the dashboard.

Add to your Claude Desktop config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "intuition": {
      "command": "node",
      "args": ["/path/to/Intuition-/lib/mcp/server.js"]
    }
  }
}

getTrustScore(address: string)

• Returns aggregated trust score and breakdown

getAttestations(filters: object)

• Query attestations with optional filters

verifyCredential(address: string, claim: string)

• Verify specific credentials for an address

findTrustedExperts(topic: string, limit: number)

• Discover highly trusted addresses in a domain

# Get trust score
GET /api/trust-score?address=0x...

# Query attestations
GET /api/attestations?subject=0x...

# Execute MCP tool
POST /api/mcp
{
  "tool": "getTrustScore",
  "params": { "address": "0x..." }
}

• Chain ID: 1155 (Intuition Mainnet)
• RPC: https://rpc.intuition.systems/http
• Explorer: https://explorer.intuition.systems
• GraphQL: https://mainnet.intuition.sh/v1/graphql (Mainnet data)

Note: The GraphQL API indexes mainnet attestation data. The blockchain operates on mainnet (Chain 1155), and attestation queries return mainnet results.

/app
  /api              # API routes
  /dashboard        # Interactive UI
  /docs             # Documentation
/lib
  /intuition        # Intuition client
  /mcp              # MCP server logic
/components
  /ui               # UI components

The application is deployed at: https://www.intuitionmcp.xyz

Deploy your own:

vercel --prod

• AI-powered credential verification
• Trust-gated content access
• Reputation-based filtering
• Expert discovery systems
• Decentralized identity verification

The trust score calculation uses:

• Predicate Weighting: Hardcoded weights for common predicates (verified=1.5x, trusted=1.3x, expert=1.2x)
• Normalization: Scores normalized as percentage of total attestations
• Positive/Negative Signals: Tracks both endorsements and warnings

Transitive Trust Computation

• Implement relative trust scores using graph traversal
• Integration with EAS Transitive Trust SDK
• Read more: Transitive Trust Model

Graph Indexer

• Current GraphQL limitations prevent deep graph traversal
• Future: Spark-based indexer for full trust graph analysis
• Enable computation of network effects and trust propagation

Governance-Based Weights

• Move from hardcoded to community-governed predicate weights
• Implement whitelist of high-quality predicates
• Dynamic weight adjustment based on network consensus

Minimum Stake Threshold

• Filter attestations below minimum TRUST token stake
• Configurable threshold (currently MVP without filtering)
• Normalize by user's total position rather than absolute stake

This project was built for the Intuition Ecosystem Grant program.

• Live Demo
• Video demo
• Intuition Portal
• Intuition Documentation
• MCP Documentation

MIT License

Built with Next.js, TypeScript, and the Model Context Protocol.