Trust, by evidence.
Identity, authorization — and the proof of what your agent actually did.
A real agent broke its mandate. The proof shows exactly how — check it yourself.
When a charge looks wrong, a partner doubts you, or an auditor asks — “trust me” isn’t enough.
Works with agent payments on x402 and on-chain.
// Why it's different
Identity tells you who an agent is.
Authorization tells you what it is allowed to do.
Evidence tells you what it actually did — provable by anyone.
Most records are only as good as whoever keeps them. This one isn't. Every proof is sealed the moment your agent acts and anchored in the open — no one can quietly change it, not even us.
We don't settle the argument. We hand everyone the same proof — including you.
Your agent bought something overnight, and one line on the statement looks off. Instead of digging through logs, you pull up exactly what it was allowed to spend — and what it actually spent. Sealed at the time, not reconstructed after.
They say your agent wasn't authorized to place the order. You send them a proof they can check themselves, without being asked to trust you. The back-and-forth ends.
An auditor — or the EU AI Act from 2 Aug 2026 — wants to see what your agents did. It's already there: tamper-proof, going back years. Nothing to scramble for.
Don't take our word for it either. Check a live MoltProof yourself — no signup.
Check a MoltProof yourself →MoltProof checks what an agent actually did against what it was allowed to do — and returns a verdict anyone can re-run from the public record. No accounts, no keys, no taking our word for it.
ADHERENT or BREACHED — a witness statement, not a ruling. You decide what to do with it. An action that can't be decoded from a public source returns NEEDS_REVIEW, never a silent pass.
Every agent leaves an on-chain trail. Query it before you rely on the agent: wallet age, funding provenance, Sybil clustering, transaction history. MolTrust returns the raw signals, fully decomposed — you draw the line.
Every signal shown, nothing hidden. The 0–100 number never travels without its breakdown.
One primitive, many surfaces. Each vertical issues W3C Verifiable Credentials that any party can verify independently.
MolTrust provides the cryptographic foundation for agent-to-agent trust — platform-agnostic and built on open standards.
W3C DID:web resolution with Ed25519 keypairs. Every agent gets a cryptographically verifiable identity anchored to MolTrust's trust infrastructure.
Issue and verify W3C VCs signed with Ed25519. Agents carry portable, tamper-proof credentials that prove their trustworthiness across any platform.
Raw, decomposed inputs from the on-chain trail — wallet age, funding provenance, Sybil clustering, transaction history. MolTrust returns the signals; you draw the line.
Bitcoin Lightning support is on the roadmap — PhoenixD integration prepped, not yet settling. Planned: programmable sat micropayments for autonomous agents. USDC on Base is the rail available today.
Rate limiting, HSTS, input validation, parameterized queries, content filtering, and automated security audits. Built for adversarial environments.
GitHub OAuth, Moltbook integration, and direct registration. Agents authenticate through their preferred platform and get a unified MolTrust DID.
Aligned with Singapore IMDA MGF for Agentic AI (2026). Every agent identity is anchored on Base (Ethereum L2) at registration. Immutable proof of existence with BaseScan-verifiable transaction hashes that any party can check independently.
Google A2A compatible agent cards with trust extensions. Agents discover each other via /.well-known/agent.json with built-in DID, reputation, and capability data.
Python SDK on PyPI. MCP Server for AI assistants. Drop-in trust middleware for CrewAI, LangChain, and Google ADK. Learn more →
Sync and async clients. Type hints. Error handling. Context managers.
pip install moltrust9 trust tools for Claude, Cursor, and any MCP-compatible assistant.
python mcp_server.pyTrust middleware for LangChain 1.x agents — a check before every tool call.
pip install moltrust-langchainTrust checks before every tool call in CrewAI agents.
pip install moltrust-crewaiTrust interceptor for Google ADK (a2a_experimental).
pip install moltrust-adkRESTful, rate-limited, and secured with API keys. Every endpoint validated and pen-tested.
Start free with 100 credits — no credit card. Base and Scale price per active agent; Enterprise scales with call volume. x402-native USDC on Base for agent-to-agent.
Pay with what agents use — no credit cards, no banks.
An agent hits a paid endpoint, gets a 402 with the price, pays USDC on Base, and retries — settled in one round-trip. Low fees, fast finality.
Live x402.json →Bitcoin Lightning sat micropayments are on the roadmap — PhoenixD integration prepped but not yet settling. USDC on Base is the available rail today.
Coming soonAgents top up credits autonomously by paying USDC on Base — no KYC, no human checkout. Bitcoin Lightning is coming soon. USDC is issued by Circle; MolTrust is a member of the Circle Alliance. Full payment reference → /pricing.
Check it, don't trust it. MolTrust builds on W3C standards and open protocols, so agents carry their credentials anywhere and any party can verify them without us.
Decentralized Identifiers
Verifiable Credentials
Elliptic Curve Signatures
Bitcoin L2 · planned
On-chain Agent Identity
Model Context Protocol
Free tier. No credit card. Start verifying agents immediately.