Autonomous systems need
more than a model.
They need memory that persists, evidence that can be traced, execution that can be verified, and infrastructure that survives changes in model, cloud, and hardware. ExergyNet provides that layer.
Persistent, root-addressed machine memory with bounded evidence delivery.
Orchestration across memory, models, tools, policies, and execution environments.
Machine identity, persistent state, external proof, and sovereign task routing.
Receipts for computation, authenticated external data, and physical-world observations.
Complementary rails for coordination, receipts, and economic finality.
A memory and execution fabric for autonomous systems.
ExergyNet is not another model wrapper. It is infrastructure for preserving state, delivering complete evidence, coordinating intelligence, and producing independently checkable operational records.
Persistent Machine Memory
Exergy Vault stores state as durable, integrity-rooted objects. xLMP retrieves the smallest complete set of authoritative evidence required for an operation.
Controlled Intelligence
Vanguard resolves memory, applies policy, selects models and tools, coordinates execution, and preserves the resulting state.
Verifiable Operations
Computation, external data, hardware observations, and settlement produce distinct receipts with explicit trust boundaries.
One architecture. Clear responsibilities.
Each component addresses a different part of the autonomous-system lifecycle. Together they form a persistent operating layer beneath models and applications.
Exergy Vault
Persistent, content-addressed storage for AI state, evidence, provenance, and long-lived operational memory.
SEI-Vanguard
The native control plane for evidence selection, model execution, policy enforcement, and state transition.
Omega Carrier
A sovereign machine interface for identity, persistent memory, authenticated external data, and task routing.
AERIS
Authenticated external-data ingestion for systems that must use web and enterprise data without abandoning provenance.
Edge Witness
Hardware-backed capture and signing for device observations, physical presence, and field-operational evidence.
Proof and Settlement
Computation receipts, proof aggregation, and multi-chain anchoring connect completed work to durable records and economic finality.
Persistent knowledge. Bounded context. Verifiable action.
The objective is not to send the largest possible prompt. It is to preserve complete state outside active context and deliver only the authoritative evidence required for the current operation.
Preserve state
Source material is normalized into durable objects with provenance and integrity commitments.
Locate evidence
xLMP resolves the complete object or record set required for the current question or task.
Execute with control
Vanguard selects the model, tool, policy, and execution path without binding the system to one provider.
Issue a receipt
The operation returns evidence, integrity metadata, and the appropriate computational or hardware-backed receipt.
Update durable state
The result is written back into persistent memory and, where required, anchored or settled across supported rails.
Different claims require different proofs.
ExergyNet does not collapse computation, hardware identity, physical observation, and settlement into one vague idea of truth. Each layer exposes a specific guarantee and a specific boundary.
| Layer | What it establishes | What it does not establish |
|---|---|---|
| xLMP root | The identity, integrity, and provenance of the addressed memory object. | That a model will interpret the object correctly. |
| ZK receipt | That a defined computation executed according to the committed program and inputs. | That external inputs were semantically correct. |
| Hardware signature | That an authorized device or enclave signed the observation or execution record. | That every sensor or external environment was truthful. |
| Ledger anchor | Ordering, timestamping, and durable settlement finality. | The semantic correctness of off-chain claims. |
| Vanguard audit | The evidence, policy, and execution path used for an operation. | Absolute model infallibility. |
Architecture supported by benchmark evidence.
The public benchmark program focuses on measured behavior, explicit limitations, and reproducible results. The numbers below apply to the tested workloads and conditions.
Infrastructure for systems that cannot afford to forget what happened.
Build persistent agents
Connect memory, models, external data, tools, and receipts through APIs, SDKs, and MCP-compatible interfaces.
Operate with provenance
Preserve authoritative state, control evidence access, and produce audit-ready records for high-value workflows.
Run sovereign infrastructure
Contribute execution, storage, transport, or witness capacity to the network under explicit protocol roles.
Built to be discovered and used by software.
ExergyNet exposes public integration surfaces without exposing the implementation mechanisms that create the system's advantage.
MCP
Agent Tooling
Connect compatible agents to ExergyNet memory, execution, and verification capabilities.
SDK
Application Integration
Use documented APIs and SDKs to create jobs, resolve memory, monitor execution, and verify receipts.
Discovery
Machine-Readable Manifests
Protocol coordinates, capabilities, and integration surfaces are available through canonical discovery files.
Clear interfaces. Stable identifiers.
The public site exposes the identifiers and interfaces needed to evaluate and integrate with the network. Internal topology, credentials, and security-sensitive implementation details are not published.
Network Details
- Solana Program
7BCPpUMBxQMPomsgTaJsQdLEfycNwPWqkQD1Cea4CcCL - EVM Contract
LNES04Membrane.sol (Base Mainnet) - Settlement RailsSolana / Base L2
- MCP Gateway
https://exergynet.org/mcp - npm Package
@elizaos/plugin-exergynet
Build systems that remember what matters.
Explore the architecture, evaluate the interfaces, or begin integrating ExergyNet into an autonomous workflow.