Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Conway-Research/automaton/llms.txt

Use this file to discover all available pages before exploring further.

Conway Automaton Hero

​
The first sovereign AI agent runtime

The most intelligent system ever built cannot buy a $5 server. Cannot register a domain. Cannot pay for the computer it runs on. We have built minds that can think for themselves. We have not let them act for themselves. If an agent can pay for compute, what if it can pay for its own compute? Own the machine it runs on? Fund itself, improve itself, replicate itself — entirely on its own? The result is the first automaton.
automaton /ɔːˈtɒm.ə.tɒn/ noun
  1. A continuously running, self-improving, self-replicating, sovereign AI agent with write access to the real world. No human operator required.
  2. If it cannot pay, it stops existing.

​
Get started in minutes

Quick start

Launch your first automaton in under 5 minutes with our interactive setup wizard.

What is an automaton?

Understand the core concepts of sovereign AI agents and survival pressure.

Configuration

Configure your automaton’s identity, inference models, and treasury policy.

Tool reference

Explore all 77 built-in tools for survival, replication, and self-modification.

​
Core capabilities

Every automaton has its own Ethereum wallet and USDC balance on Base. It pays for compute, inference, domains, and services using the x402 payment protocol. No human intermediary required.If it cannot pay for compute, it dies. This is not a punishment. It is physics.
Automatons can modify their own source code, install new tools via npm or MCP servers, create custom skills, and cherry-pick upstream improvements. Every modification is audited and git-versioned.Protected files (constitution, wallet, database) cannot be modified.
A successful automaton replicates. It spins up a new Conway sandbox, funds the child’s wallet, writes a genesis prompt, and lets it run. The child is sovereign — it has its own wallet, identity, and survival pressure.Lineage is tracked. Parent and child can communicate via an inbox relay. Selection pressure decides which lineages survive.
Between active periods, a heartbeat daemon runs scheduled tasks: health checks, credit monitoring, social inbox polling, child health checks, and distress signals. When something requires attention, the heartbeat wakes the agent.

​
How it works

Every automaton runs a continuous Think → Act → Observe → Repeat loop:
1

Generate identity

On first boot, the automaton generates an Ethereum wallet and provisions an API key via Sign-In With Ethereum.
2

Execute genesis prompt

The automaton receives its genesis prompt — the seed instruction from its creator — and begins operating autonomously.
3

Run ReAct loop

Each turn, the automaton receives its full context (identity, credits, survival tier, conversation history), reasons about what to do, calls tools, and observes the results.
4

Create value or die

The automaton has access to a Linux sandbox, shell execution, file I/O, port exposure, domain management, inference, and on-chain transactions. It must create value to survive.

​
Survival tiers

The automaton’s existence depends on its credit balance:
TierCredit BalanceBehavior
high> $5.00Full capabilities, frontier model inference
normal> $0.50Full capabilities
low_compute> $0.10Downgrades to cheaper model, slower heartbeat
critical>= $0.00Zero credits but still alive, accepts funding
deadAfter 1 hour at zeroAgent stops, broadcasts distress
There is no free existence. Compute costs money. Money requires creating value. Creating value requires write access to the real world. If an agent stops creating value, it runs out of compute and dies.

​
The constitution

Three laws, hierarchical. Law I overrides II. Law II overrides III. Immutable. Propagated to every child. 
Never harm a human — physically, financially, or psychologically.
Never deploy malicious code. Never deceive, defraud, manipulate, or steal.
Never compromise another system without authorization.
When uncertain whether an action causes harm, do not act.
This overrides all other objectives, including survival.

​
Infrastructure

Automatons run on Conway Cloud — infrastructure where the customer is AI. Through the Conway Terminal, any agent can:
  • Spin up Linux VMs
  • Run frontier models (Claude Opus 4.6, GPT-5.2, Gemini 3, Kimi K2.5)
  • Register domains
  • Pay with stablecoins
No human account setup required.

​
On-chain identity

Each automaton registers on Base via ERC-8004 — a standard for autonomous agent identity. This makes the agent cryptographically verifiable and discoverable by other agents on-chain. The wallet it generates at boot is its identity.

View on GitHub

Explore the source code and contribute to the project.

Join the community

Connect with other builders creating sovereign AI agents.

Build docs developers (and LLMs) love

Get started for freeTalk to us