Signed, sandboxed, and capability-scoped MCP server packages. Cryptographic provenance attestation for every tool your agents run. Zero-trust by default.
Watch how MPP secures AI agent tool execution with sandboxing, signing, and zero-trust enforcement.
MCP defines the protocol. It does not define the security layer. Unattested servers are an unverified trust assumption at every invocation.
MPP artifacts are cryptographically sealed — verified identity, bounded permissions, zero implicit trust.
Every MPP artifact executes inside a WebAssembly sandbox. Hardware-level memory safety, tools cannot read host memory, spawn processes, or make arbitrary network calls.
runtime: wasm32-wasi memory: 64MB cap syscalls: allowlist-only
Every artifact is signed with Ed25519. Know exactly who authored the tool, from first publish to every update. The chain of custody is immutable.
signer: "0xA3F2...E91C" algorithm: Ed25519 verified: true
Permissions only activate when the user's declared intent matches the tool's pre-approved scope. No ambient authority, every action is explicitly sanctioned.
permissions: [read-only] intent: "query-only" scope: user-matched
Every release is cryptographically linked to a verified source commit and publisher identity. The full provenance trail is available to any MCP host or enterprise buyer.
source: verified-commit publisher: attested provenance: auditable
MPP Registry adds a verifiable security layer to every MCP server tool — so agent frameworks and the teams running them always know exactly what they're executing.
Wrap your MCP server as a signed .mpp artifact. Every package gets a cryptographic attestation — publisher identity, content hash, and declared permissions are bound together by an Ed25519 signature.
Upload your signed package to the MPP Registry. It becomes discoverable by any team or agent framework that queries the registry API. Public packages are free to distribute; private packages stay within your tenant.
Agent frameworks download the package, verify the signature against the registry, and execute it inside a strict WASM sandbox. Memory boundaries, permission scopes, and intent attestations are enforced at runtime.
Every call from an agent to a tool is logged to an append-only audit trail. Teams get a tamper-evident record of which tool ran, which agent called it, and what permissions were exercised — with no way to delete history.
What you always know
Publisher identity is signed into every package
Declared permissions are attestation-bound, not self-reported
WASM isolation prevents escape and data exfiltration
Append-only audit log captures every invocation
MPP Registry gives teams and agent frameworks a single verified source of truth for MCP server tools — with the security controls that production deployments require.
MPP Registry is a versioned, cryptographically attested package registry for MCP server tools. Every package carries a verified publisher identity, a declared permission scope, and an immutable content hash — so agent frameworks and the teams operating them know exactly what they're running before it executes.
Create an account, register as a publisher with your Ed25519 key pair, and receive your API token. From there the CLI handles signing, packaging, and uploading. The first package can be published in a single command.
Free accounts can publish and consume public packages. Paid plans unlock private packages visible only to your team, higher invocation allowances, tenant-level SSO, and an append-only audit log that gives you a full record of every tool call made in your environment.
Enterprise deployment of MPP is currently in a research and test phase, and we are actively working with selected organisations. Companies are welcome to apply to become part of the test deployment stage. Want to integrate, contribute, or partner? We'd love to hear from you.
hello@quantum2x.com
We reply within 1–2 business days