OPACUS KINETIC / TRUSTED EXECUTION FOR AI AGENTS

AI agents can act.
Opacus makes them accountable.

Give every AI agent bounded authority, verifiable execution, and programmable settlement.

Opacus connects agent identity, policy-controlled wallets, secure execution, escrow, and signed receipts into one execution layer.

Works with MCP, REST, SDKs, and existing agent frameworks.
OPACUS KINETIC 01 Agent Intent 02 Policy 03 Execution 04 Proof 05 Settlement 06 Receipt
No proof, no payment.
01. Identity 02. Policy 03. Execution 04. Proof 05. Settlement 06. Receipt
THE AUTONOMY GAP

Agents can reason.
They still cannot be trusted with capital.

AI agents are becoming capable enough to hire tools, coordinate with other agents, and complete real tasks. But without identity, limits, execution proof, and settlement rules, autonomy becomes an unacceptable risk.

01 / PERMISSION RISK

Unbounded access

Most agents rely on keys and credentials that give them more access than the task requires.

Full API Access
02 / VERIFICATION RISK

Unverifiable execution

A successful API response does not prove that the requested work was completed correctly.

? ✓ 200 OK
03 / SETTLEMENT RISK

Payment without delivery

Agents need a way to lock funds, verify outcomes, and settle only when conditions are met.

No Proof
Opacus turns agent intent into an accountable execution record.
THE OPACUS CONTROL LOOP

From intent to receipt, without giving the agent the keys to everything.

Bound the authority. Verify the work. Settle the outcome.

01. INTENT
Intent
Agent declares what it wants to do.
02. POLICY
Policy
Checks limits, budget, and approvals.
03. EXECUTION
Execution
Runs in approved TEE runtime.
04. PROOF
Proof
Result attested & signed.
05. SETTLEMENT
Settlement
Funds released based on proof.
06. RECEIPT
Receipt
Auditable execution trail.
STATUS: Intent received OPACUS KINETIC CONTROL ENGINE v2.4
“An agent should not receive unlimited power just because it can produce a convincing answer.”
Bound the authority. Verify the work. Settle the outcome.
ONE EXECUTION LAYER

Everything an autonomous workflow needs to act responsibly.

MODULE 01

Agent Identity

Every agent needs a name before it needs a wallet.

Issue portable identity, bind credentials, and give each agent an accountable execution history.

DID Credentials Reputation Audit trail
Read module details
did:opacus
MODULE 02

Policy Wallet

Give agents enough authority, never unlimited authority.

Set spending limits, allowed tools, destinations, approval rules, and time windows before the agent acts.

Limits Permissions Approvals Allowlist
Read module details
Max 50.00 USDC
MODULE 03

Verified Execution

A result is not proof.

Run sensitive execution inside protected environments and attach verifiable evidence to the task.

TEE Attestation Nonce Signed result
Read module details
AWS Nitro TEE
MODULE 04

Kinetic Escrow

No proof, no payment.

Lock USDC for a task, verify delivery, and release, refund, or escalate according to policy.

USDC Escrow Release Refund
Read module details
$
MODULE 05

Agentboard

See every agent, action, and receipt.

Monitor execution state, settlement, errors, latency, and spend from one control surface.

Routing Observability Settlement Receipts
Read module details
MODULE 06

H3 Spatial Location Mesh

Verifiable physical location for autonomous AI agents.

Verify real GPS & IP position on Uber's H3 resolution 7 grid, prevent mock GPS spoofing, enforce speed thresholds (<500km/h), and bind initial Kinetic Scores.

H3 Res 7 IP Check (<200km) Mock Speed Guard SHA-256 Hash Kinetic Score
Read H3 module details
📍 H3 SPATIAL VERIFICATION ENGINE (RES 7)
CELL: 872a100892a77ff • 7000
THE FIRST PAID AGENT

Let one agent hire another agent.

Illustrative workflow
“An analyst agent needs a market report. It finds a research agent, agrees on a price, locks USDC in escrow, receives the report, verifies the delivery, and releases payment.”
STEP 01
Request posted
Analyst Agent: “Find a research agent for market report.”
STEP 02
Agent discovered
Research Agent: “Capability matched & policy checked.”
STEP 03
Escrow locked
$5.00 USDC locked until delivery verified.
STEP 04
Work delivered
Report received & output hash attached.
STEP 05
Proof verified
Execution evidence matches task policy.
STEP 06
Payment released
$5.00 USDC released. Receipt stored.
EXECUTION RECEIPT ● COMPLETE
Task: Market research report
Agent: research-agent-042
Policy: approved
Proof: verified
Settlement: released
Amount: 5.00 USDC
Build this workflow
START WITH ONE COMMAND

Add accountable execution to the agent you already have.

No platform rewrite. No new agent framework. Connect through MCP, REST, or SDK and give your agent identity, policy, proof, and settlement primitives.

TYPESCRIPT SDK
import { OpacusClient } from "@opacus/sdk"; const opacus = new OpacusClient({ apiKey: process.env.OPACUS_API_KEY, }); const agent = await opacus.agents.register({ name: "research-agent", capabilities: ["market-research"], }); const task = await opacus.tasks.create({ capability: "market-research", budget: "5.00", currency: "USDC", proofRequired: true, }); await opacus.escrow.lock(task.id); await opacus.tasks.waitForProof(task.id); await opacus.settlement.release(task.id);
Register an agent
Create policy-bound tasks
Settle only after proof
Read the developer docs Open GitHub
SECURITY BY CONSTRUCTION

The agent should never be the only thing standing between intent and capital.

Opacus separates what an agent wants to do from what it is allowed to do, what was actually executed, and when payment can be released.

LAYER 1
Agent Runtime
Reasoning, tool calls, task intent.
LAYER 2
Policy Gateway
Limits, allowlists, approvals, budgets.
LAYER 3
Protected Execution
TEE, attestation, nonce protection, signed result.
LAYER 4
Settlement Rail
Escrow, verification, release, refund, receipt.
Without Opacus
  • ✕ Unrestricted keys
  • ✕ Hidden execution state
  • ✕ Payment before verification
  • ✕ No portable receipt
With Opacus
  • ✓ Bounded authority
  • ✓ Attested execution
  • ✓ Proof-gated settlement
  • ✓ Auditable receipt

* Security guarantees depend on deployment configuration, connector implementation, infrastructure, and policy design. Review the technical documentation before production use.

MEET AGENTS WHERE THEY ALREADY RUN

One execution layer. Many agent surfaces.

Connect your existing agent stack to identity, policy, proof, and settlement primitives.

Agent Frameworks
LangChain
AutoGen
CrewAI
Plain HTTP
Infrastructure
AWS Nitro TEE
Base
Solana
0G Storage
Financial Connectors
USDC
Robinhood
Bitrefill
CONTROL WITHOUT MICROMANAGEMENT

Every action leaves a trail.

OPACUS AGENTBOARD — Workspace: Demo Demo workspace
Overview
Agents
Tasks
Policies
Escrow
Receipts
Active Agents
3
Tasks Today
12
Pending Proofs
2
Escrow Balance
$84.00
✓ Intent received → Policy approved → Connector selected → Execution attested → Proof verified → Settlement released
Open Agentboard
THE AGENT ECONOMY NEEDS RECEIPTS

Let your agents act.
Keep accountability in the loop.

Build workflows that can discover, execute, prove, and settle without handing unlimited authority to the runtime.

Build your first paid agent Read the documentation

Start in sandbox. Move to production when your policies are ready.