SDK and Personalized AI Agents
Download PDFIn Phase 3, SATI will provide an SDK through which third-party apps can integrate an embedded wallet, consent manager, Data Credentials, reward distributor, campaign interface and settlement gateway. Installing the SDK does not automatically make an app's data the property of SATI; the roles of the app operator, the user and SATI Labs, as well as processing purposes, legal bases and the scope of data movement, are separated through contracts and consent policies.
| Tier | Scope |
|---|---|
| Sandbox | Pseudonymous and test data only; no real rewards are generated |
| Standard | General data schemas and limited credential issuance |
| Verified | Extended capabilities for apps that have passed security, privacy and operational reviews |
| Restricted | Categories requiring separate contracts and regulatory review, such as finance, health and high-risk execution |
SATI AI agents are not intended to be a centralized super-app with unlimited read access to all raw data. They request purpose-specific credentials and user-approved context on a least-privilege basis, and access rights expire when the permitted period ends. Agent permissions are divided into five levels according to risk.
| Level | Permitted Actions | Default Controls |
|---|---|---|
| Read | Query permitted context and credentials | Limits on purpose, duration and fields |
| Recommend | Provide recommendations, analysis and alternatives | No execution; rationale shown |
| Prepare | Draft messages, bookings, purchases and workflows | User review required |
| Execute — Low Risk | Perform approved, low-risk recurring tasks | Limits, allowlists, cancellable |
| Execute — High Risk | High-risk tasks such as money, health, legal matters and account changes | Explicit per-action confirmation, or prohibited |
Agents keep audit logs of data access, tool calls, communication with external agents, proposals, user approvals and actual execution. MCP provides a design perspective for tool and data connectivity, A2A for agent-to-agent communication, and the ERC-8004 Draft for discovery, reputation and validation. SATI reviews the maturity and security of each specification and tests compatibility in stages.