Intent Layer

Authority Infrastructure for Autonomous Agents

Overview

Intent Layer is a verification-first authority layer that lets agents act autonomously within explicitly approved, bounded constraints.

Instead of giving agents raw keys or broad permissions, Intent Layer separates authority approval from action execution:

  1. Humans approve bounded authority (policies)
  2. Agents propose actions as signed intents, verified on-chain against policy constraints
  3. Only verified intents can be executed by authorized executors

This separation of authority, decision, verification, and execution makes agent autonomy testable, auditable, and safer by design.

Intent Layer is built for teams giving agents real economic power who need authority to be bounded, observable, and replaceable — without delegating keys.

Intent Layer is not a wallet.Intent Layer is not a relayer.It is the authority boundary that sits between them.

Core Model

Cloud
Policies configured
↓
Intent
Agent signs
↓
Verify
IntentManager + policy constraints
↓
Execute
Authorized executor
↓
Complete
Receipts

Two-Phase Execution

Intent Layer splits action execution into two explicit phases — after authority has been approved:

Phase 1Verification
  • An intent is submitted to the IntentManager
  • On-chain policy constraints are evaluated
  • If valid, an execution event is emitted
  • If invalid, execution halts permanently
Phase 2Execution
  • An authorized executor picks up the validated intent
  • Operations are reconstructed from emitted calldata
  • Transactions are executed on-chain
  • Completion is emitted and observable

Nothing executes unless it has already passed verification.

Policy-Driven Guardrails

Policies define what may execute, where, when, and how much.

Budget capsAllowed venues / protocolsFunction selectorsTime windowsDelegated signer scopes

Policies are enforced before execution, not during or after.

Executor Portability

Executors never have custody and never have discretion.

Any authorized executor can execute a verified intent — and apps can switch executors without migrating wallets or rewriting policies.

This prevents relayer lock-in and enables an execution market over time.

Intent Layer Cloud

Intent Layer Cloud is the authority lifecycle and explanation layer.

It coordinates:

  • Policy creation and versioning
  • Execution retries and durability
  • Receipt generation and audit trails
  • Budget tracking and observability

Cloud never holds custody or execution power. It is an authority interpreter, not a custodian.

What's Live Today (v1)

LIVE

Networks

Local Hardhat / AnvilBase Sepolia (public demo)

Core Protocol

  • Policy-enabled smart wallets
  • Multi-policy per wallet architecture
  • IntentManager (on-chain verification)
  • Two-phase execution (verify → execute)
  • Delegated signing (agents / relayers)
  • V2 + V3 constraint system (Merkle leaves + policy headers)
  • TOCTOU protection via header binding
  • Executor-agnostic architecture

SDK + MCP

  • Wallet creation (counterfactual + deploy)
  • Policy creation with presets (trading, treasury, operations)
  • Agent clients: transfer(), swap(), supply(), withdraw()
  • Multi-operation bundles
  • Cloud-backed runtime (ielSdk.cloud.connect)
  • MCP server for AI agent integration
  • Typed error handling
  • Coupled and decoupled execution modes

Intent Layer Cloud

  • Policy lifecycle API
  • Execution history + receipts
  • Budget tracking and observability
  • Agent-readable error responses
  • Execution retries and durability

UI Demo

  • Wallet + policy presets
  • Successful execution paths
  • Constraint failures (budget, venue, selector)
  • Agent attempts blocked by policy
  • Two-phase lifecycle visibility

Validation

E2E tests covering:

TransfersSwapsAave supply / withdrawExecutor portability (decoupled execution)

Explicitly Out of Scope (v1)

OUT OF SCOPE

This is intentional. The following are out of scope for v1 and should not be assumed:

Cross-chain routingFully permissionless executor networksOn-chain storage of full policy artifactsPrivacy-preserving executionMEV optimizationToken-based incentivesStrategy optimizationWorkflow orchestrationAgent discovery or ranking

These are future extensions or permanent non-goals, not blockers to protocol validity.

What's Coming Next

PLANNED

SDK Extensions

  • Wallet discovery (listWallets(address))
  • Wallet + policy recovery / rehydration
  • Policy version history access
  • Relayer SDK packaging (hardening + observability)

Intent Layer Cloud Extensions

  • Policy request workflow (agents request authority expansion)
  • Enterprise attestations
  • Notification channels for approval requests
  • Template-based approvals (class-based authority grants)

Mental Model (TL;DR)

Humans approve authority.

Agents decide actions.

Executors perform. Cloud observes.

Wallets remain safe.

Intent Layer makes agent autonomy safe, portable, and verifiable — without delegating keys or trusting opaque relayers.