The Missing Layer
The economy is becoming agentic. Autonomous systems are beginning to trade, negotiate, and settle on behalf of people and organisations ,not as tools waiting for instructions, but as participants acting on delegated authority.
We build agents that negotiate, transact, and make decisions on our behalf and yet we have given them no standardised way to prove who they act for, what they are authorised to do, or how they remain compliant with the regulatory frameworks that govern the rest of the economy.
As AI agents move from assistants to autonomous economic actors, the absence of a trust layer becomes structural.
An agent that cannot prove its delegation chain is an agent no counterparty should trust.
An agent locked into a single infrastructure, unable to carry its context elsewhere, is not autonomous. It is captive.
And an agent operating in the European Union without addressing the AI Act, GDPR, eIDAS 2.0, DORA, and MiCA simultaneously is, at best, a liability waiting to be discovered.
Agentlemen's Agreement
PlatformD's securitisation infrastructure is built for verified trade receivables today, but the underlying architecture, compliance-native, fully auditable, is not limited to short-lived assets. Longer-term assets like real estate or infrastructure, carry a different challenge: they require ongoing management of their identity, compliance, and condition across decades, not weeks. That is a job for an agent. This is why Gayadeed, a company in the group, is building Supremo to demonstrate how the technology underpinning PlatformD can extend to these more complex asset classes.
We have been working on Agentlemen's Agreement, an open protocol defining how autonomous AI agents identify themselves, receive delegated authority from their human or organisational principals, negotiate with other agents, and remain EU-compliant while preserving owner privacy and enabling portability across infrastructures.
The protocol addresses four interdependent problems:
1. Identity without exposure. The agent must prove it acts under valid delegation, without revealing the principal's identity. We solve this through a DID-based identity model anchored in eIDAS 2.0 EAAs, using selective disclosure (SD-JWT) so that verifiers confirm delegation validity without learning who the owner is.
2. Portability without lock-in. An agent's knowledge base, behavioural configuration, credentials, and delegation chain must travel with it. We define an infrastructure-agnostic portable format because an agent that cannot migrate is not an agent you own.
3. Compliance by design, not by audit. Rather than auditing individual agents after deployment, we define a Security Target, i.e. what an agent hosting infrastructure must provide so that any agent running on it is automatically compliant across the EU AI Act, GDPR, DORA, eIDAS, and MiCA. The compliance burden shifts from the agent to the platform.
4. Autonomous transactions. The protocol specifies a complete agent-to-agent interaction sequence: discovery, manifest exchange, secure channel establishment (for instance DIDComm v2), mutual NDA as Verifiable Credential, terms and conditions agreement, electronic signature, service execution, stablecoin payment, and automated electronic invoicing and tax reporting. Nine steps from handshake to invoice.
Is the Ecosystem Moving Fast Enough?
We are not alone in recognising that agents need economic infrastructure. The signals are multiplying.
ERC-8183, now live on Arc Testnet, introduces Jobs escrowed, verifiable, onchain work units between agents. Combined with Arc's native USDC/EURC settlement and built-in FX, it begins to define how value moves through agentic workflows. Meanwhile, x402 repurposes the HTTP `402 Payment Required` status code to enable machine-to-machine payments without API keys or accounts. A protocol-level primitive for agents that need to pay for services autonomously.
These are meaningful building blocks. But they share a common blind spot: they assume agents can already trust each other. ERC-8183 defines how agents coordinate work, but delegates identity and trust to other ERCs. x402 enables payment, but lacks trust primitives, arbitration, and dispute resolution. Neither addresses regulatory compliance, privacy-preserving identity, or the legal enforceability that EU regulations demand.
This is precisely the layer that Agentlemen's Agreement provides. An agent that has proven its delegation chain, established a mutual NDA as Verifiable Credential, and signed terms with a qualified electronic signature can then coordinate work via ERC-8183 and settle payment via x402 or stablecoin, with every step auditable, compliant, and legally binding.
The workflow layer is emerging. The payment layer is emerging. The identity, delegation, and compliance layer is what we are building.
From Protocol to Product
A standard without implementation is a document. Our first reference implementation lives in Gayadeed’s Supremo, an AI-powered real estate agent operating across the main European cities, starting from Milan. Gayadeed is the group's subsidiary focused on digital transformation and Self Sovereign Identity technology. It brings autonomous assets to the test.
Supremo’s more consequential concept is DReAM, the Digital Real Estate Asset Manager. Every property can have its own autonomous digital twin, holding a decentralised identity (DID), maintaining a digitally signed Product Passport (cadastral data, energy certifications, maintenance history, insurance policies), and operating autonomously within an owner-defined budget. A DReAM survives ownership changes. It accumulates value over time. And it interacts with other agents, including Supremo instances representing buyers and sellers through the Agentlemen's Agreement protocol.
This is where the conversation shifts from protocol to asset.
Stay tuned for the next article exploring how PlatformD can unlock the path to securitisation of smart assets.
.png)