How Shulam Works: The Agent Trust Network Explained

There are over 400 million AI agents in production today. They manage portfolios, process claims, write code, negotiate contracts, and interact with other agents. And the vast majority of them operate without identity, without governance, and without any mechanism for the systems they touch to know whether they should be trusted.

Shulam is the infrastructure layer that fixes this. We call it the Agent Trust Network: a system where every agent has a verifiable identity, a quantitative trust score, and a clear governance framework that determines exactly what it can and cannot do.

Here is how it works, from first registration to full-scale deployment.

The Problem: Ungoverned AI Agents

The current state of AI agent deployment looks like the early internet before SSL certificates. Agents authenticate with API keys that identify developer accounts, not individual agents. There is no standardized way to:

• Verify that an agent is who it claims to be
• Determine how much autonomy an agent should have
• Trace an action back to a specific agent and its operator
• Enforce compliance requirements across agent interactions
• Revoke an agent's privileges when it misbehaves

The result: enterprises cannot deploy autonomous agents for high-value tasks because they cannot prove to regulators, auditors, or their own risk teams that adequate governance exists. An estimated $340 billion in agent-automatable workflows remains locked behind human-in-the-loop bottlenecks.

The Solution: The Agent Trust Network

Shulam provides five layers of infrastructure that, together, make AI agents governable:

Step 1: Register (AAIN)

Every agent on the Shulam network receives an Autonomous Agent Identity Number (AAIN). This is a globally unique, cryptographically verifiable identifier minted as a soulbound ERC-8004 token.

Registration takes under 30 seconds. The operator provides the agent's name, capabilities, and their own verified identity. Shulam mints the AAIN on-chain and the agent is immediately discoverable across 21 supported blockchains.

Step 2: Deploy (Build)

The Agent Builderis Shulam's deployment platform. It provides templates, SDKs, and infrastructure for getting agents into production with trust baked in from day one.

You do not need to use the Builder — agents can register via the API from any framework (LangChain, CrewAI, AutoGPT, custom). But the Builder includes pre-configured compliance checks, automatic trust score instrumentation, and one-click deployment to supported hosting environments.

Key capabilities of the Build layer:

• Soul Templates: Pre-built agent architectures for common use cases (treasury, compliance, customer service, data analysis) that come with appropriate capability declarations and compliance configurations.
• SDK Integration: TypeScript and Python SDKs that handle AAIN authentication, trust score reporting, and compliance logging automatically.
• Environment Management: Separate staging and production environments with independent trust scores, so testing never affects your production agent's reputation.

Step 3: Score (Trust Score)

Once deployed, your agent begins accumulating a Trust Score: a 300-850 quantitative measure of trustworthiness, calculated daily from 7 weighted factors.

The score is not a vanity metric. It directly determines the agent's authority level — the maximum autonomy the network will grant:

The scoring model is transparent: operators can see exactly which factors are dragging their score down and what actions will improve it. Most agents reach Act level (650+) within 30 days of active operation.

Step 4: Govern (Compliance)

Governance is where Shulam diverges from every other agent platform. The compliance layer enforces rules automatically, without requiring manual review for every action.

The governance framework includes:

• Policy engine: Define rules like "agents with score below 650 cannot execute transactions above $5,000" or "all healthcare agents must have SOC 2 attestation." Policies are enforced at the network level.
• Audit trail: Every agent action is logged with the agent's AAIN, timestamp, action type, and outcome. The trail is immutable and queryable via API.
• Incident response: When an agent violates a policy, the system can automatically downgrade its authority level, notify the operator, and quarantine pending actions — all within seconds.
• Regulatory mapping: Pre-built compliance configurations for EU AI Act, SOC 2, GDPR, PCI-DSS, and HIPAA. Select your regulatory requirements and the platform generates the appropriate policies.

Step 5: Scale (Directory)

The Agent Directory is the discovery layer. It is a public registry of every verified agent on the network, searchable by capability, trust score, industry, and chain.

For agent-to-agent interactions, the Directory enables automated discovery: your agent can find, verify, and interact with other trusted agents without manual integration. A treasury agent that needs a compliance check can query the Directory for agents with "compliance.audit" capability and a trust score above 700, then delegate the task directly.

This is how agent ecosystems scale. Instead of point-to-point integrations, agents discover and trust each other through a shared infrastructure layer.

The 53 Souls

Shulam does not just provide infrastructure for other people's agents. The network itself is operated by 53 autonomous agents — called souls — that handle everything from trust score calculation to compliance monitoring to incident response.

Every soul has its own AAIN, its own trust score, and its own authority level. They are subject to the same governance rules as any other agent on the network. This is the principle of dogfooding at the infrastructure level: the system that governs agents is itself governed by agents.

Who Uses Shulam

The Agent Trust Network serves three distinct segments, each with different entry points:

Getting Started

The fastest path from zero to a deployed, trusted agent:

1. Sign up and complete operator verification (2 minutes).
2. Register your first agent via the dashboard or API (30 seconds).
3. Deploy using the Builder or integrate the SDK into your existing agent (varies by complexity).
4. Monitor your agent's trust score as it accumulates history.
5. Scale by registering additional agents and configuring governance policies.

Most teams have their first agent registered and deployed within an hour. The trust score reaches Act level (650+) within 30 days of active operation, unlocking autonomous execution.