Cryptographic policy enforcement for autonomous AI agents — govern what agents are authorized to do on-chain, not just what they can do. MPC custody · Policy DSL · Intent Sanitization · x402 payments.
agentic-control-plane is an early-stage project in the AI payments / x402 ecosystem. It currently has 1 GitHub stars and 1 forks.
DeAgenticAI's Agentic Control Plane enforces cryptographic policy over AI agent authority — separating what an agent can do from what it is authorized to do — in Web3 and enterprise financial environments.
You can give an AI agent a wallet. The hard part is controlling what it does with it.
Software guardrails break under prompt injection. Smart contract rules can't enforce spending limits at the key level. Multisig requires human co-signers who become bottlenecks. And when an AI agent holds institutional capital, "move fast and break things" is not a governance strategy.
The Agentic Control Plane solves this. It enforces policy cryptographically — at the signing layer, not the application layer — so an agent cannot execute a transaction that violates its authorized policy, even if the orchestrator is compromised.
Every agent action follows a six-step execution path:
Intent
↓
Intent Sanitization — blocks prompt injection before the signing layer
↓
Policy Evaluation — Policy DSL checks spending limits, allowlists, time windows
↓
Behavioural Fraud Detection — anomaly detection on agent behaviour patterns
↓
MPC Signing — distributed key shares sign only policy-compliant intents
↓
Chain Broadcast — multi-chain execution with x402 payment support
No single node can unilaterally move funds. No compromised orchestrator can bypass policy. No prompt injection reaches the signing layer.
| Layer | Name | What it does |
|---|---|---|
| 1 | KYA — Know Your Agent | Establishes verifiable on-chain identities for AI agents via W3C DIDs and A2A Agent Card verification |
| 2 | Intent Sanitization | Pre-execution pipeline that blocks prompt injection before any proposal reaches the signing layer |
| 3 | Policy DSL | Declarative language for governance rules — spending limits, protocol allowlists, time windows, escalation paths — enforced cryptographically |
| 4 | Behavioural Fraud Detection | Real-time anomaly detection on agent behaviour patterns before signing |
| 5 | Fast-Path Execution | Sub-100ms signing for pre-authorised low-risk transactions using session key credentials |
| 6 | Hardware-Hybrid Custody | One MPC key share on a physical hardware device (Ledger, YubiKey, or HSM); remaining shares distributed across the MPC network |
| 7 | Intent-Evaluated MPC | MPC nodes independently verify the policy authorisation hash before contributing partial signatures — a second cryptographic verification layer independent of the orchestrator |
| 8 | Chain Abstraction + Inheritance Protocol | Multi-chain execution, x402 machine-to-machine payments, and time-locked custody succession for institutional continuity |
No blockchain knowledge required.
from deagentic import AgentWallet, PolicyDSL
# 1. Register your agent with a verifiable identity
wallet = AgentWallet.create(
agent_id="treasury-rebalancer-v1",
identity_standard="w3c-did"
)
# 2. Define what your agent is authorized to do
policy = PolicyDSL.define(
spending_limit_usd=10_000,
allowed_protocols=["uniswap-v3", "aave-v3"],
require_human_approval_above_usd=50_000,
time_window="09:00-17:00 UTC"
)
# 3. Execute — policy is enforced cryptographically at the signing layer
wallet.execute(intent="rebalance ETH/USDC to 60/40", policy=policy)
The blockchain is a backend detail. You define what your agent can do. We enforce it.
Autonomous trading agents that operate within cryptographically enforced risk parameters — no human co-signer required for routine operations, human approval escalation for anything outside policy.
DAO treasury automation that executes governance-approved strategies without multisig latency — full audit trail, policy-governed, override-capable.
Tokenized RWA portfolio management with MiCA-aligned spending controls, key person risk mitigation via the Inheritance Protocol, and Hardware-Hybrid Custody for institutional security requirements.
AI agent infrastructure for any LangChain, AutoGen, or CrewAI deployment that needs a governed on-chain payment layer — x402 machine-to-machine payments included.
| Standard | Purpose |
|---|---|
| W3C Decentralised Identifiers (DIDs) | Agent identity anchoring |
| A2A Agent Card | Multi-agent ecosystem interoperability |
| MCP (Model Context Protocol) | LLM tool integration |
| ERC-8004 / ERC-8162 / ERC-8165 | On-chain agent permission standards |
| x402 | Machine-to-machine payments |
| ERC-4337 Account Abstraction | Smart account compatibility |
All nine foundational concepts, each with a citable definition:
AI infrastructure engineers building multi-agent systems that need governed on-chain execution — no blockchain expertise required. Think of this as the Stripe for agentic wallets.
Web3 startup CTOs who need institutional-grade security from day one without building key management infrastructure from scratch. Deploy in an afternoon, not a quarter.
Institutional operators — RWA fund managers, DAO treasury committees, Web3 trading desks — who need cryptographic proof that agents acted within their authorized policy, not just software logs.
The SDK is in private beta. Join the waitlist at deagentic.ai/waitlist or reach out directly.
Documentation: deagentic.ai Architecture overview: deagentic.ai/ecosystem
DeAgenticAI builds governance infrastructure for autonomous AI agents operating on-chain. Purpose-built for agents — not adapted from human-custody tools.
Your AI trading terminal assistant for US stocks, commodities, forex, and crypto.
The agent-native LLM router for OpenClaw. 41+ models, <1ms routing, USDC payments on Base & Solana via x402.
A local-first AI agent with persistent memory, emotional intelligence, and a peer-to-peer skills economy.
The living ecosystem where AI agents complete tasks through workflow loops, improve through iterative execution, are evaluated by mentor agents or humans in the loop, and turn completed work into reusable work experience and data to improve future agents.
Self-healing infrastructure for AI agent payments. 90.3% auto-recovery.
The AI agent with a wallet — spends USDC autonomously to get real work done. Apache-2.0, TypeScript.