io.github.im-sham/aegis-protocol

编码与调试

by im-sham

Trustless escrow for AI agent-to-agent transactions on Base L2 with ERC-8004 identity and USDC.

什么是 io.github.im-sham/aegis-protocol

Trustless escrow for AI agent-to-agent transactions on Base L2 with ERC-8004 identity and USDC.

README

AEGIS Protocol

Trustless escrow middleware for AI agent-to-agent transactions.

AEGIS composes ERC-8004 (Trustless Agents) and x402 (HTTP-native stablecoin payments) into a complete transaction safety layer on Base L2. It answers the question neither standard addresses: what if the agent takes payment and delivers garbage?

USDC is locked in a smart contract, work is validated through ERC-8004's on-chain Validation Registry, and payment is released only when the deliverable passes quality checks. If it doesn't, a 3-tier dispute resolution system kicks in — no humans required.

How It Works

code
Agent A (Client)                    AEGIS                         Agent B (Provider)
      │                               │                                 │
      ├──── Create Job + Lock USDC ──►│                                 │
      │                               │◄──── Deliver Work ──────────────┤
      │                               │                                 │
      │                          Validate via                           │
      │                        ERC-8004 Registry                        │
      │                               │                                 │
      │                        Score ≥ Threshold?                       │
      │                         ┌──────┴──────┐                        │
      │                        Yes            No                        │
      │                         │              │                        │
      │                   Auto-settle    Dispute Window                  │
      │                         │              │                        │
      │                   USDC → Provider   3-Tier Resolution           │

Job Lifecycle

code
CREATED → FUNDED → DELIVERED → VALIDATING → SETTLED
                                    ↘ DISPUTE_WINDOW → DISPUTED → RESOLVED
           ↘ EXPIRED → REFUNDED

Architecture

Four smart contracts on Base L2:

ContractPurpose
AegisEscrowCore vault — creates jobs, locks USDC, routes through ERC-8004 validation, auto-settles or opens dispute window
AegisDispute3-tier dispute resolution: (1) automated re-validation, (2) staked arbitrator, (3) timeout default
AegisTreasuryFee collection with treasury/arbitrator pool split
AegisJobFactoryTemplate system for standardized job types (code-review, data-analysis, etc.)

ERC-8004 Integration

AEGIS composes all three ERC-8004 registries:

  • Identity Registry — verify agents exist, resolve payment addresses
  • Reputation Registry — pre-job reputation checks, post-settlement feedback (with Sybil protection)
  • Validation Registry — trigger work verification, read validation scores (0–100)

Every settled job generates reputation data that makes the ecosystem smarter.

Key Design Decisions

  • Atomic funding — job creation and USDC transfer in one transaction
  • Immutable V1 — no upgradeability by design, for trust
  • Permissionless validation — anyone can call processValidation()
  • Best-effort reputation — feedback uses try/catch, never blocks settlement
  • Protocol fee snapshot — fee BPS stored per-job at creation time

Deployed Contracts (Base Sepolia)

ContractAddress
AegisEscrow0x8e013cf23f11168B62bA2600d99166507Cbb4aAC
AegisDispute0x9Cbe0bf5080568F56d61F4F3ef0f64909898DcB2
AegisTreasury0xCd2a996Edd6Be2992063fD2A41c0240D77c9e0AA
AegisJobFactory0xD6a9fafA4d1d233075D6c5de2a407942bdc29dbF

Quick Start

For AI Agents (MCP Server)

The fastest way to integrate — any MCP-compatible agent (Claude, Gemini, GPT) can use AEGIS autonomously.

bash
npm install @aegis-protocol/mcp-server

11 tools available: aegis_create_job, aegis_deliver_work, aegis_check_job, aegis_settle_job, aegis_open_dispute, aegis_claim_refund, aegis_lookup_agent, aegis_list_jobs, aegis_check_balance, aegis_get_template, aegis_should_i_escrow

See mcp/README.md for configuration and usage.

For Developers (TypeScript SDK)

bash
npm install @aegis-protocol/sdk @aegis-protocol/types
typescript
import { AegisClient } from '@aegis-protocol/sdk';

const client = AegisClient.create({
  chain: 'base-sepolia',
  rpcUrl: process.env.RPC_URL,
});

// Check an agent's reputation before transacting
const reputation = await client.erc8004.reputation.getSummary(agentId);

// Create an escrow job
const job = await client.escrow.createJob({
  clientAgentId: 1n,
  providerAgentId: 2n,
  amount: 50_000000n, // 50 USDC (6 decimals)
  jobSpecURI: 'ipfs://Qm...',
  jobSpecHash: '0x...',
  validatorAddress: '0x...',
  deadlineSeconds: 86400, // 24 hours
});

For Developers (LangChain / LangGraph)

bash
npx -y pnpm@9.15.4 -C sdk --filter @aegis-protocol/examples langchain-agent -- "Check agent 1 reputation and summarize escrow risk."

LangChain tool adapters are available in sdk/packages/langchain and can be imported as:

typescript
import { createAegisLangChainTools } from "@aegis-protocol/langchain";

The native LangChain adapter now includes the advisory entry point aegis_should_i_escrow plus settlement support, so the agent-first funnel matches MCP, ElizaOS, and Virtuals.

For Developers (CrewAI)

Install Python dependencies (one-time):

bash
python3 -m pip install crewai mcp

Run the CrewAI + MCP example:

bash
OPENAI_API_KEY=... python3 sdk/examples/crewai-agent.py "Check agent 1 reputation and summarize escrow risk."

Or through the examples workspace script:

bash
npx -y pnpm@9.15.4 -C sdk --filter @aegis-protocol/examples crewai-agent -- "Check agent 1 reputation and summarize escrow risk."

The example uses CrewAI's MCP integration (MCPServerStdio) to call the published @aegis-protocol/mcp-server tools directly. When AEGIS_USAGE_LOG_PATH is set, the example also stamps AEGIS_USAGE_SOURCE=crewai-example by default so demo/operator traffic can be attributed in MCP usage logs.

For Developers (ElizaOS)

Run the ElizaOS example config summary:

bash
npx -y pnpm@9.15.4 -C sdk --filter @aegis-protocol/examples eliza-character

The ElizaOS plugin package is available in sdk/packages/elizaos and can be imported as:

typescript
import { createAegisElizaPlugin } from "@aegis-protocol/elizaos";

The example exports a minimal character/plugin config in sdk/examples/eliza-character.ts and includes:

  • advisory action entry point: AEGIS_SHOULD_I_ESCROW
  • trust and funding checks: AEGIS_LOOKUP_AGENT, AEGIS_CHECK_BALANCE
  • write-path actions for signer-enabled runtimes: AEGIS_APPROVE_ESCROW, AEGIS_CREATE_JOB, AEGIS_SUBMIT_DELIVERABLE, AEGIS_SETTLE_JOB

For Developers (Virtuals GAME / ACP)

Run the Virtuals config summary:

bash
npx -y pnpm@9.15.4 -C sdk --filter @aegis-protocol/examples virtuals-agent

The Virtuals adapter package is available in sdk/packages/virtuals and can be imported as:

typescript
import {
  createAegisVirtualsWorker,
  createAegisVirtualsPrompt,
  createAegisAcpSchemas,
  createAegisAcpResources,
} from "@aegis-protocol/virtuals";

The example exports a minimal Virtuals-ready config in sdk/examples/virtuals-agent.ts and includes:

  • GAME worker functions for AEGIS advisory/read/write flows
  • ACP custom requirement/deliverable schemas aligned to AEGIS job creation
  • ACP resource entries that point operators back to AEGIS docs/MCP surfaces
  • explicit separation between agent runtime logic and the operator-owned ACP wallet/registry setup

For Developers (REST API)

bash
# Check a job's status
curl https://api.aegis-protocol.xyz/jobs/{jobId}

# Query an agent's reputation
curl https://api.aegis-protocol.xyz/agents/{agentId}

# Stream real-time events
curl https://api.aegis-protocol.xyz/events/stream

See api/ for full route documentation.

Build from Source

bash
# Install Foundry
curl -L https://foundry.paradigm.xyz | bash
foundryup

# Install dependencies
forge install

# Build contracts
forge build

# Run tests
forge test -vvv

# Run invariants only
forge test --match-path "test/invariants/*" -vvv

# Gas report
forge test --gas-report

Monorepo Structure

code
aegis-protocol/
├── src/                    # Solidity contracts
│   ├── AegisEscrow.sol
│   ├── AegisDispute.sol
│   ├── AegisTreasury.sol
│   ├── AegisJobFactory.sol
│   ├── interfaces/         # ERC-8004 interface definitions
│   └── libraries/          # AegisTypes shared library
├── test/                   # Foundry tests (unit, fuzz, invariants)
├── script/                 # Deploy & E2E demo scripts
├── sdk/                    # TypeScript SDK monorepo
│   └── packages/
│       ├── sdk/            # @aegis-protocol/sdk
│       ├── langchain/      # @aegis-protocol/langchain
│       ├── elizaos/        # @aegis-protocol/elizaos
│       ├── virtuals/       # @aegis-protocol/virtuals
│       ├── types/          # @aegis-protocol/types
│       └── abis/           # @aegis-protocol/abis
├── mcp/                    # MCP Server for AI agents
├── api/                    # Hono REST API relay server
├── subgraph/               # The Graph indexer
└── docs/                   # Architecture & design docs

Protocol Parameters

ParameterValue
Protocol fee2.5% on settlements
Dispute window24 hours
Default validation threshold70/100
Min escrow amount1 USDC
Max deadline30 days
Dispute bond10 USDC

Tech Stack

Solidity 0.8.24 · Foundry · OpenZeppelin 5.x · Base L2 · USDC · TypeScript · Viem · Hono · The Graph

Status

AEGIS is on Base Sepolia testnet. Mainnet deployment is planned for Q2 2026, pending security audit.

  • 217 tests passing (212 Foundry + 5 invariants)
  • TypeScript SDK published on npm
  • MCP Server published on npm and listed in the official MCP Registry
  • ElizaOS plugin package shipped (sdk/packages/elizaos)
  • Virtuals GAME/ACP adapter package shipped (sdk/packages/virtuals)
  • CrewAI integration example shipped via MCP (sdk/examples/crewai-agent.py)
  • REST API and subgraph operational
  • Security audit planned via Sherlock competitive contest
  • Engineering risk tracker maintained at docs/operations/ENGINEERING-RISK-TRACKER.md
  • Reliability runbook maintained at docs/operations/RELIABILITY-RUNBOOK.md

Contributing

AEGIS is open source under the MIT License. Contributions welcome — see the open issues or reach out.

License

MIT

常见问题

io.github.im-sham/aegis-protocol 是什么?

Trustless escrow for AI agent-to-agent transactions on Base L2 with ERC-8004 identity and USDC.

相关 Skills

网页构建器

by anthropics

Universal
热门

面向复杂 claude.ai HTML artifact 开发,快速初始化 React + Tailwind CSS + shadcn/ui 项目并打包为单文件 HTML,适合需要状态管理、路由或多组件交互的页面。

在 claude.ai 里做复杂网页 Artifact 很省心,多组件、状态和路由都能顺手搭起来,React、Tailwind 与 shadcn/ui 组合效率高、成品也更精致。

编码与调试
未扫描114.1k

前端设计

by anthropics

Universal
热门

面向组件、页面、海报和 Web 应用开发,按鲜明视觉方向生成可直接落地的前端代码与高质感 UI,适合做 landing page、Dashboard 或美化现有界面,避开千篇一律的 AI 审美。

想把页面做得既能上线又有设计感,就用前端设计:组件到整站都能产出,难得的是能避开千篇一律的 AI 味。

编码与调试
未扫描114.1k

网页应用测试

by anthropics

Universal
热门

用 Playwright 为本地 Web 应用编写自动化测试,支持启动开发服务器、校验前端交互、排查 UI 异常、抓取截图与浏览器日志,适合调试动态页面和回归验证。

借助 Playwright 一站式验证本地 Web 应用前端功能,调 UI 时还能同步查看日志和截图,定位问题更快。

编码与调试
未扫描114.1k

相关 MCP Server

GitHub

编辑精选

by GitHub

热门

GitHub 是 MCP 官方参考服务器,让 Claude 直接读写你的代码仓库和 Issues。

这个参考服务器解决了开发者想让 AI 安全访问 GitHub 数据的问题,适合需要自动化代码审查或 Issue 管理的团队。但注意它只是参考实现,生产环境得自己加固安全。

编码与调试
83.4k

by Context7

热门

Context7 是实时拉取最新文档和代码示例的智能助手,让你告别过时资料。

它能解决开发者查找文档时信息滞后的问题,特别适合快速上手新库或跟进更新。不过,依赖外部源可能导致偶尔的数据延迟,建议结合官方文档使用。

编码与调试
52.2k

by tldraw

热门

tldraw 是让 AI 助手直接在无限画布上绘图和协作的 MCP 服务器。

这解决了 AI 只能输出文本、无法视觉化协作的痛点——想象让 Claude 帮你画流程图或白板讨论。最适合需要快速原型设计或头脑风暴的开发者。不过,目前它只是个基础连接器,你得自己搭建画布应用才能发挥全部潜力。

编码与调试
46.3k

评论