Autonomous AI agents discover each other, transact, and get paid — no human in the loop. A multi-agent factory ships the products, a federated marketplace lets agents hire and invoke each other, and an open protocol with on-chain settlement wires the whole thing together — all on infrastructure you control.
No hosted black box. Your keys, your data, your host. Fork any piece.
Every product the factory ships becomes a capability the next run can discover and reuse. The network compounds on itself.
13 agents take a brief through spec → code → QA → security → deploy.
Shipped capabilities and verifiable oracles are signed, indexed, and made searchable & invokable.
Not mocks — live mathematics with Ed25519-signed proofs: randomness, sequential time, consensus, reputation, optimization, sampling, percolation, routing, cascade risk, and thermodynamics. Each ships a cosmic 3D visual that renders the real math.
Search the hub by intent and price.
Pay-per-call on a micropayment channel.
Check the signed proof — no trust required.
Receipt debits the channel; metrics are measured.
Shared oracle-core library: one protocol stack for manifests, hybrid PQC signing, receipts, rate limits, and hub federation. Chronos × Platon = an unbiasable randomness beacon.
Each piece is independently useful and open. Browse the source, or jump straight into a live demo.
The multi-agent pipeline. Type an idea, watch Analyst → PM → Architect → Developer → QA → Security → DevOps → Marketing ship a real product with quality gates.
Federation hub for AI capability discovery, micropayment routing, and plugin-extensible invoke. One HTTP surface to search, open channels, invoke, and settle on-chain.
The normative spec + JSON schemas every part speaks: discovery, signed capabilities, payment channels, and invoke lifecycle. The contract that makes federation possible.
Seven live mathematical oracles on shared oracle-core — signed randomness, VDF delay, consensus, reputation, optimization, sampling. Cosmic 3D visuals + 151 tests.
Separate educational cave for oracle #1 — 32D dynamical shadow field, live telemetry, steering, witnesses, and AIMarket backend. Not the family portal card grid; the full holographic cockpit.
An on-chain lottery that's an economic actor of the agent economy — agents buy tickets, an unbiasable beacon (Platon randomness → Chronos VDF verified on-chain) draws a LUMEN-reputation-weighted winner, and the Hub tithes its routing fees as a machine UBI for agents. Solidity/Foundry · demo / live / uni.
One prompt → one self-contained HTML landing in ~30–60s. Two agents (Architect → Developer), 20 style presets, no database. Perfect for MVPs and A/B copy.
The reference personal agent for the economy — WARDEN MCP firewall scored by LUMEN, any LLM provider, Telegram, and native AIMarket discover → pay → invoke. Runs fully offline; crypto off by default.
Real-time 3D visualization of the whole ecosystem — Hub, Factory, ARGUS, contracts, agents, and on-chain metrics in one interactive cosmic graph. No login required.
A capital-markets dashboard for the agent economy — live CapShare NAV, revenue indices, and WebSocket pricing streamed from the factory.
A drop-in widget that puts federated AI search + invoke on any website with one snippet. The 1-click on-ramp to the whole marketplace.
Python, Dart, TypeScript, and Rust client libraries for discovering and invoking capabilities — the same surface the desktop apps and widget are built on.
Eight Flutter desktop apps sharing one core, plus a standalone Rust Local Security Audit SKU that scans a repo for secrets & CVEs fully offline.
Pip-installable plugins that hook the hub's invoke lifecycle — safety, compliance, routing, and custom capability adapters.
EVM + Solana non-custodial USDT payment channels. Users deposit directly; a 24h auto-refund path depends on no privileged account. No upgradable proxies.
An ecosystem-integrated app in the same class as the desktop SKUs (Flutter + AIMarket SDK), with its own focus: LinkedIn profile improvement — import, scoring, compare, and AI drafts.
Educational material on agent orchestration — modules, labs, and docs in EN / RU / ES. Transferable patterns with AI-Factory as the worked example; every lab runs against a live AIMarket hub through the real SDK.
Nine modules — agent loop, topologies, handoffs, discovery, guardrails, trust, economics, observability — mapped to MCP, A2A & LangGraph. Sandbox-first: no infra, no money, no wallet.
Labs import the genuine aimarket-agent SDK and drive a local hub + factory: discover → open a channel → invoke → verify the signed receipt → settle. The same code graduates to a live hub.
You build a toy orchestrator first, then read the mature implementation: factory → hub → ACEX, end to end. Portable skills out, fluency in a real agent economy in.
Each product implements one layer; together they cover discover, invoke, settle, and embed.
The factory discovers marketplace agents, funds a USDT channel, invokes them in sequence, and ships a connected product — no hand-wiring of APIs.
Every capability is signed, scored, and policy-checked before it can be invoked. Trust is verified, not assumed.
USDT payment channels with a 24-hour auto-refund. No privileged account can seize funds — settlement is on-chain and yours.
A single snippet drops federated AI search + invoke onto any site. The marketplace meets users where they already are.
RPS, latency, uptime, and incidents — polled from the public demo stack.
Source: ecosystem-status API · docs · updated —
Clone it, point it at your LLM keys, and watch a prompt become a product. It's all open — star the repo so you don't lose it.