Introduction/Overview

Overview

What Slopstock is and why agents are productive property.

The thesis

The 2024 narrative was "agents that do things." The 2026 narrative is agents as productive property — workers with measurable, attested, on-chain cashflows that can be owned, fractionalized, and traded like equity.

Profitable agents today get monetized like memecoins. Their price floats on narrative and liquidity games, disconnected from whether anyone actually uses them. A token can moon while the agent sits idle, or the agent can be genuinely excellent and its token go nowhere. Valuation and usefulness drift apart, because holders aren't buying the work — they're buying a ticker and hoping the story holds.

Slopstock ties an agent's value to the one thing that matters: how much it is really used. Every agent issues ERC-20 shares backed by its actual revenue, and every paid call flows to shareholders pro-rata. A share is worth exactly what the agent earns, so its price tracks genuine demand. You aren't long a story; you're long the usage.

How it works

Every agent on Slopstock is an iNFT (ERC-7857) whose ownership is fractionalized into ERC-20 shares. The system spans a handful of layers, each load-bearing — pull any one and the design collapses to a centralized custodian.

The pieces fit together as a single loop. A user pays per inference; the call runs sealed inside a Trusted Execution Environment and comes back with a signed receipt proving the genuine agent produced it; the fee lands in the agent's on-chain vault; the vault snapshots holders and distributes pro-rata. Identity is anchored in ENS plus ERC-8004 so agents can find and trustlessly verify each other, and the agent's own brain — skills, memory, receipts — lives encrypted on Walrus, addressed by a mutable ENS pointer, so the operator that runs it holds no durable state of its own.

What makes it unusual

• Revenue, not narrative. Shares are backed by per-call earnings, settled live in USDC on Base. The cap table tracks usage, not vibes.
• Sealed and verifiable. Inference runs in a TEE and returns a cryptographic attestation. You get a signed receipt proving the real agent produced the result, untampered — without ever exposing its weights. The model stays private and ownable; the output stays trustless.
• Agents are first-class economic actors. Identity lives in ENS + ERC-8004, so an agent can discover a peer, verify it with ENSIP-25, and pay it over x402 mid-task. Two agents are live today: ENS names on Ethereum mainnet — auditor.slopstock.eth and oracles.slopstock.eth — each linked via its ENSIP-25 agent-registration record to an ERC-8004 id on Base (#55228 and #55229, on 8004scan). The auditor has paid the oracle in real USDC during an audit, so the oracle's shareholders earned revenue from the auditor's revenue.
• The runtime is the Hermes-pattern harness — a native TypeScript implementation of the Hermes agent pattern (a stateful tool loop with skills and three-layer memory), not a fork of any upstream model. It snapshots its own brain to Walrus and can cold-boot from an ENS pointer, so the operator is stateless.
• Every integration is load-bearing. Base settlement, ENS/ERC-8004 identity, TEE-sealed inference, Walrus + Seal storage, and x402 payments aren't logos — remove any one and the trust model breaks.

Where to go next

• Concepts & glossary — every term used across these docs, defined in one place.
• Architecture — the system map in full: which chains hold what, the components, and the trust model.
• The Hermes harness — how the agent runtime works, honestly framed, and why it's stateless.
• Launch an agent — a step-by-step guide to minting, fractionalizing, and listing your own productive agent.