Base-native financial control for autonomous agents

Give AI agents money. Not unlimited power.

Nomylax gives autonomous AI agents controlled financial authority on Base through budgets, policies, risk controls, treasury monitoring, execution rules, and verifiable audit trails. Every request is evaluated before it settles.

Policy enforcementDeterministic
SettlementVerified
NetworkBase
Owner treasury Policy ring Research $20 / day Trading $50 / day Yield $35 / day Ops $8 / day Base

Live control plane

Approved · settles
Blocked at ring
Research Scout $4.50
PurposeMarket dataset
RecipientApproved
PolicyWithin limits
Risk score12 · LOW
ApprovedSettling on Base
Built on Base Policy-enforced Agent-native Real-time risk Auditable
0
Policy checks · demo
0
Blocked actions · demo
0
Protected value · demo
0
Agents monitored · demo

The gap

Autonomous agents should not hold unrestricted wallets.

Agent frameworks have solved capability. None of them solved authority. The moment an agent holds a private key, its worst possible decision and its best possible decision cost exactly the same to execute.

Unbounded authority

An agent handed raw wallet keys inherits the full authority of its owner. One malformed loop, one prompt injection, one hallucinated recipient, and the loss is the size of the treasury rather than the size of the mistake.

Unpredictable spend

Autonomous systems retry failed calls, escalate to expensive models, and buy the same dataset four times. Spend curves that look flat in testing turn steep the moment an agent meets the real world.

No financial governance

The stack is crowded with tools for what an agent can do. Almost nothing governs what it should be financially permitted to do, with limits that hold even when the model is wrong.

The Nomylax model

One treasury.
Many agents.
Explicit financial rules.

Owner

You hold the Base wallet and the capital. Nomylax never takes custody or unrestricted control of your treasury.

Agent

Each agent requests an economic action. It states intent: amount, recipient, purpose. Then it waits for a decision.

Nomylax

Deterministic policy and live risk scoring decide whether the action is financially permissible, then execute it.

Owner treasury $24,732.68 Nomylax core Policy · Risk · Simulation Research $20 / day Max tx $10 Trading $50 / day Max tx $25 Operations $8 / day Recurring only Same treasury · different authority

How it works

A financial constitution for autonomous agents.

01

Connect

Connect a Base wallet and register the agent that will act on its behalf.

02

Define

Write the agent's financial constitution: limits, recipients, reserves, thresholds.

03

Observe

The agent submits an economic intent instead of signing a transaction directly.

04

Enforce

Deterministic policy and the risk engine inspect the action and return a decision.

05

Verify

Approved actions execute on Base and are verified and recorded after settlement.

Live decision

Watch Nomylax make the decision.

Every request runs the same sequence: policy first, then risk, then simulation. Policy is deterministic, so a model cannot argue with it. Switch the request to see what changes.

Decision trace TRACE 0x8f3a…7c2e
Agent
Research Scout
Request
$4.50
Purpose
Market dataset
Recipient
Approved
Execute SettlementBase · USDC

Capabilities

The control layer autonomous finance was missing.

Policy engine

Rules that hold when the model is wrong

Spending limits, approved recipients, asset restrictions and treasury reserves, evaluated deterministically on every request. No inference in the enforcement path.

LIMIT $25
Treasury intelligence

Balance, allocation, runway, utilisation

See what your autonomous workforce actually costs: burn per agent, capital allocated versus idle, and how many days of operation the treasury supports at current velocity.

Risk engine

A score on every action

Contract, behaviour, operational and transaction risk combined into one number the policy layer can act on.

Agent network

Many agents, one workspace

Register, budget, pause and compare every autonomous system drawing on the same treasury.

Shadow mode

Spend nothing to find out

Run the agent against live policy with simulated settlement and read the report before funding it.

Audit layer

Every decision leaves a record

Requests, approvals, rejections, risk scores and executions are written to an append-only trail with the policy version that produced each outcome.

Shadow mode

Test the agent before funding the agent.

The agent runs its real logic against real policy. Nomylax evaluates every request and records what would have happened: approvals, blocks, projected burn, all without moving a single dollar. Most agents fail their first shadow run. Better here than on Base.

Capital at risk$0.00
Policy accuracy1:1 live
24-hour shadow report Research Scout
Requested$0.00
Would approve$0.00
Would block$0.00
Policy violations0
Critical events0
Estimated monthly burn$0

Recommendation · Review before activation

One critical event and four violations in 24 hours. Tighten the maximum transaction to $10 and add the two unknown recipients to the approved list before enabling protected mode.

Financial intelligence

Protection is only half the system.

Nomylax also shows how your autonomous workforce uses capital: what each agent costs, what it returns, and how long the treasury lasts at current velocity.

Treasury performance · 7D
$24.73M
+7.42%
Treasury allocation
USDC58.3%
ETH20.1%
ETH LSTs12.7%
Other8.9%
Daily burn · per agent
$312.40 Trading 41%
Operating runway
79 days

At current velocity, excluding the $100 emergency reserve held outside agent authority.

Execution outcomes
Executed978
Blocked306
Held for review47
Expired11

Risk engine

Every financial action earns a score.

The Nomylax Risk Score, or NRS, compresses seven signals into a single value from 0 to 100. Policy thresholds act on the number, so an agent's authority contracts automatically as its behaviour drifts.

Amount anomaly18
Unknown recipient94
Contract risk22
Spend velocity67
Liquidity risk14
Policy deviation88
Operational failures31
12
/ 100 NRS
Low

Select a band to preview how the gauge reads. Thresholds are configurable per agent.

Safe mode

When behaviour changes, autonomy stops.

Three failed transactions, spend velocity above 200%, or a critical risk event moves the agent down a state. Recovery is deliberate: an owner decision, never an agent one.

State 01

Autonomous

The agent operates freely inside its constitution. Approvals settle without owner involvement.

State 02

Watch

Anomaly detected. Limits tighten, high-risk actions are held for review, the owner is notified.

State 03

Safe mode

New spending stops. Execution pauses, unknown recipients are refused, the treasury is preserved.

01Block all new spending
02Pause agent execution
03Refuse unknown recipients
04Freeze automated actions
05Alert the owner
06Preserve the reserve

Use cases

Built for the agent economy.

Base-native

Designed for agents operating on Base.

Agent commerce needs settlement cheap enough for a $0.40 API call and fast enough for a decision loop. Nomylax enforces policy in milliseconds, then settles on Base.

Ownership stays with your Base wallet
Scoped, revocable spend permissions
USDC-denominated agent operations
Low-cost settlement for micro-payments
Onchain execution with post-verification
Agent-native commerce primitives
BASE

Security architecture

Security is architecture, not a feature.

A request passes through eight boundaries before it becomes a settled transaction. Each one can stop it.

L1Owner authorityThe human holds the keys and the capital. Every permission below is delegated, scoped and revocable.
L2Base permission boundarySpend-permission integration is designed as a second boundary; the current build does not claim it is wired onchain.
L3Policy engineDeterministic evaluation of limits, recipients, assets and reserves. No model in the decision path.
L4Risk engineSeven signals scored to NRS, compared against the agent's configured threshold.
L5Transaction simulationThe call is simulated before signing. Unexpected state changes fail the request.
L6ExecutionWhen live execution is explicitly configured, approved actions enter the Base execution pipeline. Shadow Mode never settles.
L7Post-transaction verificationSettled results are checked against the intent that was approved. Divergence triggers Safe Mode.
L8Audit proofRequest, decision, policy version and outcome are written to an append-only record.

Trust center

Everything important should be verifiable.

Nomylax publishes its own operating record. If the policy layer ever fails open, the number below stops being zero, and you will see it here first.

Operational
System status
Not established
Autonomous decisions
Not established
Policy checks
Not established
Executed
Not established
Blocked
Not established
Value protected
0
Policy bypasses
$0
Unauthorised loss
Open trust center

Getting started

From wallet connection to protected autonomy.

Connecting a wallet is step one of six. It establishes ownership. It does not hand anything to an agent.

01
Connect wallet
Your Base wallet establishes ownership of the treasury.
02
Create workspace
Name the treasury and pick a default risk profile.
03
Connect agent
Bring your own agent, or start with a demo agent.
04
Write constitution
Set limits, recipients, reserves and thresholds.
05
Run shadow mode
Watch what the agent would do with no money at risk.
06
Enable protected mode
Autonomy begins, inside the boundary you defined.

Strategy & status

A control layer that complements the agent stack.

Start with bounded financial authority for individual agents, then expand through agent-framework integrations, protocol adapters, team policy administration and enterprise monitoring.

Intent gating
Product wedge
Developer → Pro → Teams
Planned packaging
Not declared
Token
Not declared
Revenue share
Not declared
Funding target
Explanation only
AI Control Copilot
Open Control Copilot Read technical documentation

Build autonomous finance with control.

Give AI agents the ability to operate without surrendering unlimited authority over capital.