Building the operating layer for on-chain intelligence

altlayer Protocol

AI infrastructure and autonomous agents designed for the decentralized web. This page covers what drives the team, how the technology works, and where the protocol is headed.

Mission

Why altlayer Exists

Blockchain data is rich. Most teams can't easily work with it.

Fast by default

Agents need responses in milliseconds, not minutes. The altlayer platform was built with latency as a first-class constraint from day one.

Open composition

Every product in the ecosystem is composable with the others. An agent registered via ERC-8004 can immediately talk to AltLLM and execute through AltClaw.

Chain-agnostic

The protocol does not pick winners between L1s and L2s. altlayer's tooling runs across Ethereum mainnet, Arbitrum, and custom rollups built on the RaaS stack.

Developer-first

Documentation, SDKs, and local testing tooling — including Hardhat plugins — ship with every major release. Builders should be productive within an afternoon.

Technology

How the Stack Is Built

Four distinct layers. Each one works independently, and together they form a complete AI-on-chain runtime.

Layer 1 — Registry

ERC-8004 Agent Standard

The ERC-8004 standard defines how autonomous agents register themselves on-chain, expose capabilities, and accept payments. Think of it as an ENS for agents — but with executable interfaces attached.

8004scan is the explorer for this registry. It lets anyone verify an agent's metadata, inspect past executions, and deploy new agents without writing a single line of Solidity.

Layer 2 — Inference

AltLLM Model Platform

AltLLM provides a unified interface to frontier models — GPT-4o, Claude 3.5 Sonnet, DeepSeek-V3, Gemini 1.5 Pro, and more than two dozen others. Credit-based pricing means you pay per token, not per seat.

The API is OpenAI-compatible, so any project already using the OpenAI SDK can switch with a one-line config change.

Layer 3 — Execution

AltClaw DeFi MCP

AltClaw implements the Model Context Protocol for DeFi. Agents that need to swap tokens, provide liquidity, or read price feeds route requests through AltClaw's private RPC cluster — which strips MEV exposure before transactions land on-chain.

Arbitrum is the primary supported L2 at launch, with additional networks being added each quarter.

Layer 4 — Infra

Rollups-as-a-Service

Teams that need dedicated throughput can spin up a custom rollup in minutes. The RaaS layer supports OP Stack, Arbitrum Orbit, and zkSync ZK Stack — with no-code deployment dashboards and managed sequencers.

Local testing is handled via the altlayer Hardhat plugin, which simulates rollup finality in a forked environment.

Approach

How the Team Ships

Short cycles. Real users at every stage.

Open specification first

Before any product ships, the team publishes a detailed spec — either as an EIP draft or a public design doc. Community feedback shapes the final API surface. This is how ERC-8004 got its current form: three public comment rounds over six weeks.

Build in public, break things early

Testnet deployments go out before mainnet by at least 30 days. Bug reports during this window receive a direct response from the engineer who wrote the code — not a ticketing system. The altlayer platform has run 14 public test cycles to date.

Audit, then ship

Every smart contract module gets an external audit before it touches real funds. Results are posted in full — including findings the team disagreed with. So far, three audit reports are publicly available in the altlayer documentation.

Iterate with real metrics

Usage data — anonymized and aggregated — feeds directly back into the product roadmap. If a feature is not being used six weeks after launch, it gets cut or rethought. The team does not maintain a graveyard of half-built features.

Team

The People Behind altlayer

Researchers, engineers, and protocol designers. Not a marketing org with a dev team attached.

Protocol Research

ERC authorship, cryptography, and post-quantum work including the altpq library and ETH2030 implementation.

Infrastructure Engineering

Sequencer design, RPC infrastructure, and the private relay network powering AltClaw's MEV protection.

AI & Agent Systems

Model integrations, the AltLLM routing layer, and agent orchestration frameworks including MCP tooling.

Developer Experience

SDKs, Hardhat plugins, documentation, and the onboarding flows that get new teams productive fast.

The team is distributed across North America, Europe, and Asia. Backed by Polychain Capital, Binance Labs, Jump, and Breyer Capital. Engineering decisions stay inside the engineering team.

Roadmap

What Comes Next

No vague promises. Specific targets, tracked publicly.

Shipped

ERC-8004 v1

On-chain agent registry with full metadata schema, payment routing, and the 8004scan explorer live on mainnet.

Shipped

AltLLM Public API

25+ models available through a single OpenAI-compatible endpoint. Credit top-ups via crypto and card.

In progress

AltClaw MCP v2

Expanded DeFi protocol coverage, cross-chain intent routing, and a public SDK for agent developers building on Arbitrum and OP Stack chains.

Planned

Post-quantum rollup layer

Integration of altpq's ML-KEM and SLH-DSA primitives into the RaaS sequencer — bringing FIPS 205-compliant signing to production rollups.

Explore the full picture

The FAQ page covers detailed questions about how the protocol works, what it costs, and whether it is right for your use case.