Metalogue
A federated platform for deploying secure AI agents across enterprise data on a zero-trust architecture — and an open standard for the way those agents talk to each other.
Metalogue
If agents are going to query each other across organisations, the query itself needs identity, attestation and interoperability — a “SWIFT for AI cognition.”
Context
Enterprises want AI agents working across their data, but not at the cost of trust boundaries. Metalogue is the platform for deploying those agents federatively — each staying inside a zero-trust perimeter — and MFQP is the protocol layer that lets them exchange queries safely.
What I built
- Authored MFQP — the Metalogue Federated Query Protocol — an open standard for secure AI-to-AI queries.
- Specified Ed25519 attestation so every query carries a verifiable, signed identity.
- Designed for interoperability with the Model Context Protocol (MCP), rather than a closed island.
- Shipped reference implementations in Python, TypeScript and Go so the standard is usable on day one.
How it's built
MFQP treats an inter-agent query the way a payments network treats a message: a signed, attested envelope with a known identity on both ends. Ed25519 signatures give cheap, fast public-key attestation; MCP compatibility means the protocol plugs into the emerging agent-tooling ecosystem instead of competing with it. Three reference SDKs — one per major backend language — keep the spec honest and prove it interoperates.
Outcome
An open, multi-language standard for federated agent communication, published with working reference implementations.