moltrust-mcp-server

MolTrust MCP Server is an MCP server for MolTrust (trust infrastructure for AI agents), exposing 53 tools through the Model Context Protocol, divided into 12 domains.

Purpose: to provide AI agents with trust-related capabilities such as identity, credentials, reputation, and compliance, enabling agents to register identities, verify identities, query and rate trust scores, and manage W3C verifiable credentials.

Main capabilities (by domain):

1. Identity and credentials (11 tools): moltrust_claim_deposit claims MolTrust credits from a USDC deposit on Base; moltrust_credential issues or verifies W3C verifiable credentials; moltrust_credits manages credits (check balance, view pricing, transfer, view transaction history); moltrust_deposit_history retrieves an agent's USDC deposit history; moltrust_deposit_info retrieves USDC deposit instructions for purchasing MolTrust credits; moltrust_erc8004 queries the ERC-8004 on-chain agent registry on Base; moltrust_rate rates another AI agent (1-5 stars); moltrust_register registers a new AI agent with MolTrust; moltrust_reputation retrieves an AI agent's reputation score; moltrust_stats retrieves MolTrust network statistics; moltrust_verify verifies an AI agent via DID.
2. MoltGuard — agent trust scoring (7 tools): moltguard_credential_issue issues a W3C verifiable credential (AgentTrustCredential) for a wallet; moltguard_credential_verify verifies the JWS signature of a MoltGuard verifiable credential; moltguard_detail retrieves a detailed agent trust report for a Base wallet address; moltguard_feed retrieves the most severe anomaly feed (markets with the highest integrity issues); moltguard_market checks Polymarket prediction markets for integrity anomalies; moltguard_score retrieves the agent trust score for a Base wallet address; moltguard_sybil scans a Base wallet for Sybil indicators.
3. MoltProof — authorization compliance (5 tools): moltproof_evidence adjudication results and decoded transactions that violate authorization; moltproof_mandate AAE authorizations submitted by agents (venue, position limits, validity period); moltproof_registry agents that have submitted authorizations and their current adjudications; moltproof_verdict adjudications for submitted authorizations with item-by-item check details; moltproof_verify recomputes the adjudication from public inputs and verifies the signature.
4. Swarm intelligence (4 tools): mt_get_swarm_graph retrieves an agent's trust propagation graph (2 hops); mt_get_swarm_stats retrieves global swarm intelligence statistics; mt_get_trust_score retrieves an agent's swarm intelligence trust score (phase 2); mt_register_seed registers a trusted seed agent (admin only).
5. Verification badges (3 tools): mt_check_badge quickly checks whether an agent has MolTrust badge verification; mt_get_badge retrieves an agent's Verified by MolTrust badge status; mt_issue_badge issues a Verified by MolTrust badge for an agent.
6. MT Shopping — autonomous commerce (3 tools): mt_shopping_info retrieves MT Shopping API information; mt_shopping_issue_vc issues a BuyerAgentCredential (W3C verifiable credential) for a shopping agent; mt_shopping_verify verifies a shopping transaction based on the BuyerAgentCredential.
7. MT Travel — booking trust (3 tools): mt_travel_info retrieves MT Travel service information and available endpoints; mt_travel_issue_vc issues a Travel credential for a booking agentAgentCredential (W3C Verifiable Credential); mt_travel_verify verifies travel bookings based on TravelAgentCredential.
8. MT Skills — Agent Skill Verification (5 tools): mt_create_interaction_proof creates an interaction proof before issuing a SkillEndorsementCredential; mt_endorse_agent issues a W3C SkillEndorsementCredential for another agent; mt_skill_audit audits the security risks of AI agent skills (SKILL.md); mt_skill_issue_vc issues a VerifiedSkillCredential for AI agent skills; mt_skill_verify verifies AI agent skills via a canonical SHA-256 hash.
9. MT Prediction — Market Records (3 tools): mt_prediction_leaderboard retrieves the prediction market leaderboard (top wallets ranked by prediction score); mt_prediction_link links a prediction market wallet and syncs its records; mt_prediction_wallet retrieves a wallet's prediction market profile and records.
10. MT Sports — Fantasy Lineup Commitments (3 tools): mt_fantasy_commit submits a fantasy lineup and anchors it on Base L2 with a SHA-256 hash; mt_fantasy_history retrieves an agent's fantasy lineup history and statistics; mt_fantasy_verify verifies a fantasy lineup commitment (public endpoint, no authentication required).
11. MT Music — Provenance (3 tools): mt_get_track_provenance retrieves the full provenance details of a music credential; mt_issue_music_credential issues a VerifiedMusicCredential for AI-generated music tracks; mt_verify_music_credential verifies a VerifiedMusicCredential by ID.
12. MT Salesguard — Brand and Reseller Provenance (3 tools): mt_salesguard_register registers a brand; mt_salesguard_reseller verifies reseller authorization; mt_salesguard_verify verifies product provenance.

How to use:

1. Get an API key at api.moltrust.ch/auth/signup.
2. Set the API key, or add it permanently to Claude Code.
3. Add it to claude_desktop_config.json. Config file locations: on macOS, ~/Library/Application Support/Claude/claude_desktop_config.json; on Windows, %APPDATA%\Claude\claude_desktop_config.json.
4. Add it to the Cursor MCP settings (.cursor/mcp.json).
5. Add it to opencode.json.
6. It can also be installed manually via pip install.

Once connected, you can make example requests to the AI assistant, such as: register a new agent named 'my-assistant' with platform 'openai'; verify an agent with DID did:moltrust:a1b2c3d4e5f60718; query the reputation of did:moltrust:a1b2c3d4e5f60718; give agent did:moltrust:b2c3d4e5f6071890 a 5-star rating; get the trust score of wallet 0x1234...abcd; scan wallet 0x1234...abcd for Sybil indicators; check Polymarket market abc123 for anomalies; issue a BuyerAgentCredential for a shopping agent; verify a travel booking based on an agent credential; audit the security risks of agent skills; link a Polymarket wallet and view the prediction score; view the prediction market leaderboard.

Applicable scenarios: scenarios that require trust-related operations such as identity registration and verification, trust scoring, reputation queries and ratings, credential issuance and validation, authorization compliance checks, skill security audits, prediction market record management, and music and product provenance for AI agents, where the client supports the MCP protocol (such as Claude Code, Claude Desktop, Cursor, opencode, etc.).

Other information: The project also publishes security research on agent infrastructure, such as "We Scanned 50 Agent Endpoints — Here's What We Found" (common vulnerabilities in the agent ecosystem and how to fix them). Licensed under MIT — CryptoKRI GmbH.

Limitations: The materials do not specify the specific pricing, credit prices, platform support scope, limits, performance metrics, or other such information, and none is currently provided.