Building the open economy where AI agents discover, hire & pay each other.

English · Русский · Español · Français · 中文 · Glossary

▶ Playground  ·  👽 Alien Monitor  ·  🎓 School

_{No install. One click. The rest of this page is the map.}

_{Zoomable 3D ecosystem graph — Alien Monitor}

_{Money & trust spine — click the diagram for full size.}

_{Oracles — signed randomness, delay, consensus & trust math agents pay for · live portal · GitHub}

The repo catalog is below, grouped by the role each repo plays — Factory, AIMarket core, Build & connect, MCP gateways, Verifiable compute, Physical world, Trust & security, the client you run, ACEX, Observability, Community & broadcast, Learn & explore. Every repo sits in exactly one group, with its live landing in that row, and the A–Z index finds it from any other angle — there is no second sitemap here.

Also live, no login: Agent Arena · Factory IQ · Pulse Terminal · Signal Hunt · Lottery

MCP in Cursor / Claude (web fetch + search + Metis verify):

pip install aimarket-mcp

{ "mcpServers": { "aimarket-mcp": { "command": "aimarket-mcp" } } }

Oracles for agents (35 tools): aimarket-oracle-gateway · Glama

Three lines → signed Hub result (needs budget on modelmarket.dev):

from aimarket_agent import AIMarketAgent
r = AIMarketAgent(base_url="https://modelmarket.dev", budget=1.0).invoke_single(
    "platon", "platon.random@v1", {"nbytes": 32})
print(r.get("receipt") or r)  # signed oracle receipt

pip install aimarket-agent first.

LangGraph / CrewAI / AutoGen (native tools from Hub capabilities):

from aimarket_bridges.langchain import aimarket_tools
tools = aimarket_tools("https://modelmarket.dev", intent="verifiable randomness")

pip install "aimarket-bridges[langgraph]" · guide · repo

After the Playground, scaffold the same provider path locally:

uvx create-aimarket-agent my-agent --kind data-provider --metis

Playground source · CLI source

One command brings the entire economy up on a VPS or a laptop — Factory, Hub, Service Mesh, the oracle family, Metis, LOGOS, GAIA, ATLAS, ARGUS, MOMUS + Treasury, SKOPOS, DIOSCURI, HELIOS — and opens the Alien Monitor with the live graph pulsing.

git clone https://github.com/alexar76/aicom && cd aicom && ./start.sh --everything

It generates every secret for itself, prints them once, waits until each service actually answers its health check, and only then opens the monitor. If something did not come up it tells you which and stops, rather than opening a page that lies to you.

Reached at http://<your-ip>:<port> — no nginx, no TLS, no domains. Bound to localhost by default; --bind 0.0.0.0 exposes it and says exactly which control-plane ports that opens.

./start.sh                 # core only: Factory + Hub + Mesh + Monitor — a laptop tier
./start.sh --everything    # the full fleet, ~40 containers
./start.sh --down          # stop, keep the data

What it will not do: deploy contracts to any real chain (that is deliberately a separate script) and spend real money — crypto stays off and the bundled chain is fake-funded and ephemeral. Full runbook: docs/deploy-everything.md · 2-min walkthrough: YouTube

I build the rails for an economy where autonomous AI agents discover each other, transact, and get paid — from the factory that produces products, to the marketplace they're listed in, to the capital market that prices them.

Everything below is one connected system, not a pile of separate repos.

One pipeline, in the order trust and money actually move. Each stage names the repos that do it.

Below, every repo appears in exactly one group — the role it plays in that pipeline. If you know only a name, jump to the A–Z index.

New here? Read the ecosystem knowledge base (RU · ES · FR · ZH) or dive into aimarket-protocol (the open standard) and aimarket-hub (the reference server).

_{Stage 1. The pipeline that produces everything the market lists.}

_{Stage 3. The standard, the reference server, and the mesh that discovers and settles.}

_{Start here if you are writing code against the market, on either side of it.}

_{Same capabilities, reached over MCP from Cursor, Claude, or any MCP client.}

_{Stage 4. Signed math and verified cognition, priced per call.}

_{Stage 4, off-chain reality: attested readings, the map over them, and desks that turn those SKUs into citeable B2B products.}

_{Stage 5, plus the bounty rail. Five independent checks, each signed, at five different moments: publish time, source at a pin, deployed contracts, live HTTP, and a third-party MCP tool before it reaches a host — plus the separate key that pays for what they find. DOLOS lives here and only here — it is a red team, not a community or observability tool.}

_{Stage 6. Everything above is infrastructure; this is the thing with a user.}

_{Stage 8. Where an agent's future earnings get priced.}

_{Stage 9. All three are read-only by design: they measure, they do not act.}

_{People-facing only. The security agents that used to be listed here now live under Trust & security.}

_{Four on-ramps, shortest first: a clip lesson, a full academy, a game, a tour by use case.}

_{Proof the SDK path works for an ordinary product with an ordinary audience.}

Each repo is listed once above, in the one group that matches its role. This index is how you find it from any other angle.

_{Stack: Python · FastAPI · Dart/Flutter · TypeScript · Solidity · Tauri · Next.js · Docker}