TrustGrade — Crypto Trust Verification

Score crypto entities on trust before acting on them. One call returns a 0–100 TrustGrade Score with letter grade, confidence, and risk tier. 500+ entities scored: DeFi protocols, stablecoins, exchanges, tokens, and AI agents.

When to use

• A user asks whether a protocol, stablecoin, exchange, or token is safe or trustworthy
• Comparing entities on trust ("USDC vs USDT vs DAI")
• An AI agent is about to execute a transaction — verify BEFORE executing
• Verifying another AI agent's identity before interacting with it
• Any crypto recommendation flow: score first, then advise

How to call (MCP)

Server: @trustgrade/mcp-server (stdio transport, no API key required)

{"mcpServers":{"trustgrade":{"command":"npx","args":["-y","@trustgrade/mcp-server"]}}}

Optional env TRUSTGRADE_API overrides the default base https://api.trustgrade.ai.

Tools

• check_trust — the primary tool: TrustGrade Score (0–100), letter grade, confidence %, risk level for any scored entity
• compare_trust — side-by-side trust comparison of multiple entities
• verify_transaction — pre-flight check before an agent executes: agent limits, business verification, trust score, recipient whitelist, manipulation flags; returns an approved/rejected decision with a cryptographic receipt
• verify_agent — verify an AI agent's identity: DID, business KYB status, trust score, active status
• get_business — business verification status behind an entity or agent
• register_agent — register a new AI agent (model, capabilities) to get scored
• list_entity_types — all supported entity types

Reading the result

• Score 0–100 with letter grade (A+ … F) and confidence %
• Risk level: LOW · MEDIUM · HIGH
• verify_transaction returns approved or rejected with reasons — a rejection is a hard stop, not a suggestion
• The score reflects evidence quality and verification status — not price prediction or investment merit

Rules

• Call check_trust before recommending any crypto entity; never call an unscored entity "safe"
• Quote the score with its grade and confidence; never round or editorialize
• If an entity isn't indexed, say so — do not guess a score
• For agent transactions: verify_transaction BEFORE execution; treat rejected as final
• Never present trust scores as investment advice or price forecasts