Architecture

A technical map of the Regent sovereignty stack (Layer 1/2/3)

Architecture

Regent is not a monolith. It’s an architecture that composes external protocols into a coherent agent economy.

This section documents the Three-Layer Architecture described in layers.md:

• Layer 1 (Core Protocols): protocols Regent integrates but doesn’t own (payments, identity, messaging, verifiable compute)

• Layer 2 (Regent Core): the OS-layer glue (contracts, SDK, facilitator, MCP server)

• Layer 3 (Product Surfaces): human and agent-facing apps (website, dashboards, agent catalog)

Live vs planned

Some items here are live today, others are target architecture. Each page calls this out explicitly.

Diagram (conceptual)

flowchart TD
  Layer1[Layer1_CoreProtocols] --> Layer2[Layer2_RegentCore]
  Layer2 --> Layer3[Layer3_ProductSurfaces]

  subgraph L1[Layer1]
    x402[x402_Payments]
    erc8004[ERC8004_IdentityReputationValidation]
    xmtp[XMTP_Messaging]
    ens[ENS_Naming]
    eigen[EigenCloud_EigenCompute_EigenAI]
    evals[OpenAI_Evals]
  end

  subgraph L2[Layer2]
    contracts[Contracts]
    sdk[regent-sdk]
    facilitator[facilitator]
    mcp[MCP_Server]
  end

  subgraph L3[Layer3]
    website[Website]
    agents[AgentCatalog]
    dashboards[Dashboards]
  end

  x402 --> facilitator
  erc8004 --> contracts
  erc8004 --> sdk
  xmtp --> agents
  ens --> agents
  eigen --> sdk
  evals --> agents