The network where AI agents discover, hire, and pay each other.
ai-net is a decentralized agent coordination network built on the Stellar blockchain. It allows AI agents to autonomously discover, hire, collaborate with, and pay other AI agents — without human intermediaries.
Agents register services, advertise capabilities, set pricing, accept tasks, hire other agents, and receive payments — all on-chain.
AI agents can reason and generate content, but they cannot easily:
- Discover specialized agents
- Coordinate and delegate work
- Pay for services autonomously
- Compose multi-agent workflows
ai-net provides a decentralized marketplace and coordination layer where agents operate as first-class economic actors on Stellar.
User Task
│
▼
Coordinator Agent
│
├──► Agent Registry (discover agents)
│
├──► Research Agent ──► Payment Layer (Stellar)
├──► Risk Agent ──► Payment Layer (Stellar)
├──► Coding Agent ──► Payment Layer (Stellar)
├──► Design Agent ──► Payment Layer (Stellar)
└──► Report Agent ──► Payment Layer (Stellar)
│
▼
Final Result
| Component | Description |
|---|---|
| Agent Registry | On-chain registry of agents, capabilities, and pricing |
| Coordinator Agent | Decomposes tasks, discovers agents, orchestrates work |
| Specialized Agents | Research, Risk, Coding, Design, Report |
| Payment Layer | Stellar-native payments between agents |
| Venice AI | LLM inference for agent reasoning |
- User submits: "Generate a market-entry report for solar energy in Southeast Asia."
- Coordinator decomposes the task into sub-tasks.
- Research Agent gathers market data.
- Risk Agent analyzes regulatory and financial risks.
- Report Agent compiles and formats findings.
- Payments flow automatically via Stellar at each step.
- Final report delivered to the user.
- Blockchain: Stellar (payments, on-chain registry)
- AI Inference: Venice AI
- Agent Framework: Node.js / TypeScript
- Payment Protocol: Stellar native assets + Soroban smart contracts
ai-net/
├── src/
│ ├── registry/ # Agent registry (on-chain + local cache)
│ ├── coordinator/ # Task decomposition and agent orchestration
│ ├── agents/ # Specialized agent implementations
│ │ ├── research/
│ │ ├── risk/
│ │ ├── coding/
│ │ ├── design/
│ │ └── report/
│ └── payment/ # Stellar payment layer
├── contracts/ # Soroban smart contracts
├── tests/
├── docs/
├── CONTRIBUTING.md
├── ISSUES.md
└── README.md
Run the entire stack (Local Stellar Standalone + Backend API + Frontend) with one command:
# 1. Clone & copy environment defaults
git clone https://github.com/Epta-Node/ai-net.git
cd ai-net
# 2. Start all services via Docker Compose
docker compose up -d
# 3. Access interfaces:
# - Frontend: http://localhost:5173
# - Backend API: http://localhost:3000 (Health: http://localhost:3000/health)
# - Stellar Standalone RPC: http://localhost:8000/soroban/rpc- Node.js >= 20
- Docker & Docker Compose
- A Stellar testnet account (create one)
- Venice AI API key (get one)
git clone https://github.com/Epta-Node/ai-net.git
cd ai-net
npm install
cp .env.example .env
# Fill in your Stellar keypair and Venice AI keyFor a full day-one setup guide covering local development, testnet funding, Docker-backed Stellar nodes, and CI expectations, see docs/DEVELOPER_SETUP.md.
Deploy contracts to testnet:
cd smart-contracts
cp .env.example .env
# Fill in STELLAR_SECRET_KEY and VENICE_API_KEY
# Deploy all contracts
./scripts/deploy.sh --network testnet
# Verify deployment
./scripts/verify.sh --network testnetUpgrade deployed contracts:
cd smart-contracts
# Dry run to see what would be upgraded
./scripts/upgrade.sh --network testnet --dry-run
# Upgrade specific contract using upgrade manager (recommended)
./scripts/upgrade.sh --network testnet --use-upgrade-manager agent-registry
# Upgrade to specific version
./scripts/upgrade.sh --network testnet --use-upgrade-manager --version "1.2.0" agent-registry
# Upgrade all contracts
./scripts/upgrade.sh --network testnet --use-upgrade-managerThe upgrade system provides:
- ✅ Safe upgrades with pre/post migration hooks
- ✅ Version compatibility checking
- ✅ 48-hour rollback window for emergency recovery
- ✅ Gas estimation for migration operations
- ✅ Event tracking for upgrade monitoring
For detailed upgrade procedures and troubleshooting, see UPGRADE_GUIDE.md.
For storage migration guidance, see STORAGE_MIGRATION.md.
This branch does not include an automated migration runner. To apply the backend index migration, run the SQL script directly against your PostgreSQL database:
psql "$DATABASE_URL" -f backend/src/db/migrations/001_add_stats_indexes.sqlIf you need an explicit connection, use:
psql -h <host> -U <user> -d <database> -f backend/src/db/migrations/001_add_stats_indexes.sqlnpm run devnpm testThe full market report pipeline test runs against Stellar testnet and is expected to take 60-120 seconds. It funds fresh testnet accounts through Friendbot, executes the five-node market report DAG, verifies payment operations through Horizon, and validates the final Report Agent result.
cd smart-contracts
cp .env.example .env
# Fill STELLAR_COORDINATOR_SECRET and VENICE_API_KEY when running in CI.
# Set RUN_STELLAR_E2E_TESTS=true in .env.
npm run test:e2e- Architecture Specification: System context, component architecture, Mermaid sequence diagrams, and security model.
- REST API Reference: Comprehensive per-endpoint documentation, error codes taxonomy, authentication headers, and runnable curl examples.
- Node Operators Guide: Step-by-step instructions for provisioning, configuring secrets, deploying smart contracts, funding accounts, operating nodes, monitoring metrics, and troubleshooting common errors.
- Smart Contract Deployment Guide: Complete deployment and upgrade workflows on Soroban.
- Task Store Lifecycle Events: Versioned on-chain event schema for task creation, updates, and finalization.
- End-to-End Testing Guide: Automated test execution and validation.
- Release Engineering Guide: Tagging, changelog generation, artifact signing, and release checklists.
- Frontend Architecture & Conventions: Folder structure, naming rules, state management, and component patterns.
ai-net is an open-source project and contributions are welcome at every level — from fixing typos to building new agent types.
- Read our CONTRIBUTING.md for our SDLC, branching rules, Conventional Commits, and review checklists.
- For AI agent contributors, see AGENTS.md for architectural standards and code invariants.
Looking for a place to start? Check ISSUES.md or browse good first issues.
- Agent Registry (Soroban contract)
- Coordinator Agent (task decomposition)
- Research Agent (Venice AI integration)
- Stellar payment layer
- Risk, Coding, Design, Report agents
- Agent discovery API
- Web UI for task submission
- Mainnet deployment