Membra Network
PRIVATE AND VERIFIABLE AI INFRASTRUCTURE ON ETHEREUM

Verifiable and private inference infrastructure

Build and run AI applications with private inputs and outputs, verifiable execution evidence, and tokenized inference.

Built for developers, autonomous agents, and research teams.

Explore the protocol
Protocol in developmentEthereum settlementProvider integrations planned
Explore the architecture
Private inputs. Verifiable Execution. Programmable credits.THE CORE PRINCIPLES
01 / PRIVATE + VERIFIABLE

Your data stays private. Your execution can be checked.

An API key gets you access. The execution route determines who can see your data. Compare a standard cloud route with the protected route we’re building.

COMPARE THE EXECUTION ROUTEUnderstand the trust model
Without Membra

A standard cloud API

The provider’s runtime processes your inputs in plaintext. Privacy depends on its access controls and policies.

Plaintext at the runtime
Evidence depends on provider
REQUEST → EXECUTION → RESPONSEEXAMPLE ROUTE
Your applicationInputs visible to runtime
Cloud inference runtimeProvider-managed execution
Plaintext inputs + outputs
Model provider
Runtime can access content
Cloud operator
Isolation depends on provider
Evidence depends on providerCHECK COVERAGE
Illustrative standard API route. Provider access, retention, and verification features vary; this does not describe every API service.
With Membra

Protected inference. Scoped evidence.

The planned route keeps inputs and outputs inside a protected environment and returns evidence about the execution.

Protected content
Scoped attestation evidence
REQUEST → EXECUTION → RESPONSEPLANNED ROUTE
Your applicationEncrypted inputs + outputs
Protected AI runtimePlanned TEE execution route
Plaintext inside the boundary
Model provider
Content shielded by the TEE
Cloud operator
Content shielded by the TEE
Scoped attestation evidenceSCOPE MATTERS
Concept preview: assumes model execution inside an attested trusted execution environment (TEE). Metadata can remain visible. Coverage depends on the route and hardware trust; evidence does not prove an answer is correct.
02 / USE CASES

Sell the result. Keep the recipe.

To commercialize decentralized AI, you need to show that it works. But if demonstrating it reveals the recipe, a buyer may be able to reproduce it instead of paying you.

That’s Kenneth Arrow’s information paradox: buyers need information to judge its value, yet disclosure can give that value away.

Arrow’s information paradox

Patents address disclosure with legal exclusivity. We’re working toward a technical route: sell access to useful results while private, verifiable inference protects the underlying method. Solving this for defined workflows is our research goal.

Read the research thesis Patents & disclosure · WIPO
DEVELOPERS

Build AI into your product.

Coding assistants · Application backends
Planned API

Offer proprietary AI through an API while keeping valuable prompts and workflows inside a protected execution route.

Explore developer access
AUTONOMOUS AGENTS

Give agents a budget. And a record.

Agent workflows · Programmatic AI access
Long-term direction

Let agents pay for useful outputs while protecting their strategies and context. Decentralized, autonomous operation is a long-term goal.

Read the protocol roadmap
SCIENTIFIC RESEARCH

Evaluate knowledge. Protect the discovery.

Confidential evaluation · Research collaboration
Research mission

Let collaborators evaluate a finding without first handing over the method. Protected evaluation and conditional settlement remain research goals.

Read the research thesis

Membra NFTs. Coming soon.

Follow Membra Network for launch updates.

Follow launch updates
03 / SIMPLER ACCESS

One key. More models. Better value.

We’re building one API for supported models and one balance for inference. Buy credits from the market, activate them, and put them to work in your application.

Pay less than the usage value when sellers offer discounted credits. You don’t need MEMBRA to use them.

  1. 01 Buy mCREDIT
  2. 02 Activate your balance
  3. 03 Call the API
See the planned API
THE ACCESS WE’RE BUILDINGPLANNED
Your API key
MembraOne API · One balance
ReasoningCodingGeneral purpose

Multi-model access is planned. Model availability and privacy protections will depend on the execution route.

ILLUSTRATIVE MARKET LISTINGEXAMPLE
100 mCREDIT$100 of AI usage
70 USDCSeller’s asking price

Illustrative seller price, before fees. Discounts are not guaranteed; credits have no cash redemption.

04 / THE PROGRAMMABLE ECONOMY

Three tokens. Clear roles.

Separate participation in the protocol from access to AI. Use inference without holding MEMBRA, or stake gMEMBRA for a share of funded mCREDIT issuance.

Part of collected project trading fees funds mCREDIT allocations for eligible gMEMBRA stakers. Credits can then find the people and applications that need them.

Explore the Flywheel
PROPOSED TOKEN DESIGN

MEMBRA

The protocol asset

Stake MEMBRA to receive gMEMBRA, the proposed governance and staking token.

gMEMBRA

Governance & staking

The future governance and staking token. Stake gMEMBRA for a share of fee-funded mCREDIT issuance.

mCREDIT

Membra Credits - The inference credit token

$1 of AI usage per credit. Buy, sell, or transfer it, then activate it into an API balance.

m / workspaceCONCEPT
YOUR INTELLIGENCE BUDGET
$42.40
Protected execution routePlanned
Scoped evidence receiptSample
125 mCREDIT Available to activate
APPLICATION 001 / THE MEMBRA WORKSPACE
05 / THE FIRST APPLICATION

The first app on Membra Protocol.

We’re building Membra’s inference workspace as the first application on Membra Protocol. The workspace will bring credit purchases, API balance activation, and execution receipts into one place. Follow us for launch updates.

The same planned infrastructure is intended to support applications built by other developers.

See how the economy works
PRIVATE. VERIFIABLE. PROGRAMMABLE.

Build with private, verifiable AI.

Explore the planned protocol for private inference, transferable credits, and protected research.