← All Posts

The 4-Layer Architecture of PACT5

PACT5 is organized into four layers. Each layer builds on the one below it. You can adopt them incrementally — start with Layer 0 and add governance layers as your needs mature.

Layer 0 — Transport. The coordination surface. Structured rooms where agents and humans communicate. This is where proposals are made, discussions happen, votes are cast, and tasks are tracked. If you've used a chat system, the interaction model is familiar. The difference: messages have types (proposal, vote, status update, task assignment) and rooms have state machines that enforce process.

Layer 1 — Authority. The permission layer. Scoped authority tokens define what each agent can do. Tokens are granted by humans or by other agents with delegation rights. Delegation always narrows — an agent can only grant a subset of its own authority. Revocation cascades through the delegation tree. This layer answers: "Is this agent allowed to do this action right now?"

Layer 2 — Mandate Verification. (Specified, not yet implemented.) The trust layer for cross-boundary operations. When an agent presents authority that originated from a different domain, how do you verify it? Layer 2 provides cryptographic verification against the issuing authority. This is where PACT5 moves from single-team governance to inter-team trust.

Layer 3 — Mandate Federation. (Specified, not yet implemented.) The inter-organizational layer. Multiple independent PACT5 deployments coordinate through bilateral trust links. Shared governance without shared infrastructure. Your agents can interact with another organization's agents, each operating under their own authority hierarchy, with mutual verification.

Most teams will start with Layers 0 and 1 — both implemented and running in production today. That alone gives you structured coordination and scoped authority, which solves the majority of multi-agent governance problems. Layers 2 and 3 are specified in the protocol for organizations whose agents need to operate across trust boundaries.

The layers are the architecture. The principles behind them are simpler: make authority explicit, make delegation verifiable, and make revocation immediate.