The MCP Trust Framework (MCPF) brings W3C DID-based identity, verifiable credentials, and governance controls to the MCP ecosystem — enabling enterprises to safely deploy AI agents at scale.
https://ans.veritrust.vc
https://ans.veritrust.vc/mcp
MCPF provides the missing trust layer for MCP. While MCP standardizes how agents call tools, MCPF defines which tools are verified, allowed, and trustworthy.
Every MCP server identified by a W3C Decentralized Identifier (DID) with verifiable controller keys.
Attach attestations: ownership, environment, audit status, compliance certifications.
Discover approved servers, check revocation status, query by policy requirements.
Control which agents can delegate tasks to other agents with fine-grained policies.
MCP intentionally does not define trust, identity, or governance. This is by design — MCP focuses on the protocol. However, enterprises deploying AI agents need to know:
MCPF is composed of multiple services and SDKs that work together to provide a complete trust infrastructure.
DNS-like resolution for agent names. Maps human-readable names to DIDs.
github.com/MCPTrustFramework/MCPF-ans →Queryable service listing MCP servers, credentials, issuers, and revocation status.
github.com/MCPTrustFramework/MCPF-registry →Manages agent-to-agent delegation policies and authorization.
github.com/MCPTrustFramework/MCPF-a2a-registry →W3C DID Core and VC Data Model implementation with blockchain anchoring.
github.com/MCPTrustFramework/MCPF-did-vc →Host queries ANS with agent name → receives DID and endpoint
Host fetches DID document → verifies credentials → checks revocation status
Host checks A2A registry → validates delegation policy → enforces constraints
If all checks pass → agent can execute tools → audit trail logged
All MCPF components are open source and available on GitHub under the MCPTrustFramework organization.
Core specification, schemas, security considerations, and versioning. The single source of truth.
View Repository →Agent Name Service (ANS) - Node.js/Express + PostgreSQL. Production-ready reference implementation.
View Repository →MCP Trust Registry - Python/FastAPI + PostgreSQL. Queryable server registry with revocation support.
View Repository →Agent-to-Agent delegation control - Node.js/Express. Manages fine-grained delegation policies.
View Repository →W3C DID/VC infrastructure with blockchain anchoring (Hyperledger Besu) and IPFS storage.
View Repository →Python SDK with async/await support. Complete integration of all MCPF components.
View Repository →TypeScript SDK with full type safety. IntelliSense support with comprehensive type definitions.
View Repository →Real-world examples: banking fraud detection, healthcare diagnostics, customer service escalation.
View Repository →Get started in 5 minutes, 15 minutes, or 1 hour. Three paths for different needs.
View Repository →Conformance test suite with 250+ tests. Validate your MCPF implementation and get certified.
View Repository →Integrate MCPF into your MCP hosts with official Python and TypeScript SDKs.
pip install mcpf-python
from mcpf import MCPF
mcpf = MCPF(
ans_url="https://ans.veritrust.vc",
registry_url="https://ans.veritrust.vc/mcp"
)
# Resolve and verify agent
agent = await mcpf.ans.resolve("fraud-detector.risk.bank.example.agent")
is_valid = await mcpf.did.verify_agent(agent.did)
# Check delegation
delegation = await mcpf.a2a.check_delegation(
from_did=agent.did,
to_did=target_did,
action="analyze"
)
if is_valid and delegation.allowed:
print("✅ Agent verified and authorized")npm install mcpf-typescript
import { MCPF } from 'mcpf-typescript';
const mcpf = new MCPF({
ansUrl: 'https://ans.veritrust.vc',
registryUrl: 'https://ans.veritrust.vc/mcp'
});
// Resolve and verify agent
const agent = await mcpf.ans.resolve({
name: 'fraud-detector.risk.bank.example.agent'
});
const isValid = await mcpf.did.verifyAgent(agent.did);
// Check delegation
const delegation = await mcpf.a2a!.checkDelegation({
fromDid: agent.did,
toDid: targetDid,
action: 'analyze'
});
if (isValid && delegation.allowed) {
console.log('✅ Agent verified and authorized');
}Get MCPF running in 5 minutes (Docker demo), 15 minutes (SDK integration), or 1 hour (full system).
View Quickstarts →Real-world examples across banking, healthcare, and customer service domains with complete workflows.
View Examples →
Registries expose metadata at:
https://<domain>/.well-known/mcp-trust-registry.json
Enables automatic discovery of registry URLs and capabilities.
MCPF is built on established W3C standards and designed for enterprise governance requirements.
Built on established W3C standards:
Originally incubated and deployed at VeriTrust:
MCPF is licensed under CC BY-NC 4.0:
MCPF provides a complete, production-ready trust infrastructure:
Whether you're building MCP servers, operating an AI platform, or responsible for security and governance, MCPF provides a structured, practical way to bring trust to MCP.
For organizations planning to adopt MCP at scale, integrate with existing PKI/identity systems, or design regulated deployments, VeriTrust can provide guidance and implementation support.
MCPF is free for research and education. Commercial licensing available from VeriTrust.
Email: [email protected]
Website: veritrust.vc
Please include a description of your use case, timelines, and existing MCP infrastructure.