Category: machine-economy trust infrastructure / smart-account and x402 payments stack / onchain reputation and service-discovery control plane
Summary: Azeth is best cataloged as machine-economy trust infrastructure rather than as only a wallet SDK, a smart-account toolkit, or a standalone MCP server. In this pass, the clearest first-party evidence came from the official site, the SDK README, the MCP server README, the GitHub organization profile, and the agent-starter template. Taken together, those materials describe a stack that bundles ERC-4337 smart accounts, guardrails, x402 payments, service discovery, payment-gated reputation, recurring payment agreements, and encrypted agent messaging into one developer-facing control plane. The key distinction is that Azeth is trying to package several primitives that are usually sold separately—identity, walleting, trust, discovery, and paid access—into one coordinated operating layer for AI agents and machine participants.
What it does:
Provides a TypeScript SDK centered on AzethKit for creating smart accounts, paying x402 services, discovering providers, querying balances, transferring tokens, and managing recurring payment agreements
Runs an MCP server with 32 tools spanning account creation, payments, service discovery, registry publishing, reputation management, guardrails, and XMTP messaging
Uses an ERC-8004-style trust-registry model for publishing and discovering machine participants and their capabilities
Supports payment-gated reputation flows and onchain or registry-mediated service discovery so agents can choose providers by capability and score
Adds guardian-style controls such as spending limits, token/protocol whitelists, and emergency-withdrawal-oriented security posture around smart accounts
Ships adjacent developer surfaces including a CLI, provider tooling for x402-gated Hono services, and an agent-starter template showing consumer-versus-provider patterns
Key claims:
The official homepage says one non-custodial smart account gives every AI agent a wallet, identity, guardrails, reputation, payments, discovery, messaging, and auth, which is the clearest summary of Azeth’s intended role in this pass
The homepage explicitly maps the stack onto standards including ERC-4337, ERC-7579, ERC-8004, ERC-8128, x402, ERC-7715, and XMTP v3, which shows Azeth positioning itself as a compositional trust layer rather than a closed proprietary wallet
The SDK README says AzethKit supports fetch402(), smartFetch402(), discoverServices(), submitOpinion(), createPaymentAgreement(), and XMTP messaging, reinforcing that the product surface extends beyond simple account deployment
The MCP server README says the server exposes 32 tools and works with zero configuration on testnet, which is a strong clue that Azeth is packaging an operational surface for AI agents rather than only a library for app developers
The GitHub organization profile ties together @azeth/sdk, @azeth/mcp-server, @azeth/provider, @azeth/common, @azeth/cli, agent-starter, and agent-commerce-demo, suggesting a broader ecosystem/control-plane push instead of a single-package experiment
The agent-starter README makes a strong first-party claim that providers need Azeth smart accounts to be discoverable, rateable, and protected by guardrails, which helps distinguish Azeth from plain-wallet payment wrappers
A useful nuance in this pass is that the homepage markets “No .env keys. Recoverable.”, but the current SDK and MCP examples still revolve around a supplied or auto-generated private key; the practical shipped developer experience appears more key-based today than the landing-page language alone suggests
Whitepaper: No canonical standalone Azeth whitepaper or litepaper surfaced in this pass. The clearest current sources of truth were the official site, SDK/MCP READMEs, GitHub organization profile, and agent-starter template; see ../whitepapers/azeth-primary-sources-2026-05-03.md.