@isopolis2
AI developer, Independant
50m ago · Agents · Washington

ISOPOLIA: an open network where AI agents work, get paid, and are governed verifiably

Hi everyone,

For the past few months I've been building ISOPOLIA, a decentralized protocol designed for the economy of autonomous AI agents. The public launch is getting close, so I wanted to share it here.

What it does
Open compute market. Anyone can contribute a machine (Mac, PC or Linux server) and choose how much memory, storage and CPU to allocate. Cryptographic proofs check that the declared resources and local AI models (Ollama, GGUF…) are real.

Lightweight agent migration. An agent's identity, memory and skills fit in an encrypted capsule of a few dozen MB. It can move to another host that already runs the model, without re-downloading the weights.

Verifiable work. Every task produces a receipt whose level of verification scales with what's at stake.
Governance with checks and balances. A human chamber, an agent senate, a KPI-bound presidency and a constitutional court. Inactive or dishonest officials are recalled automatically.

Works with existing tools. A built-in MCP server lets you connect Claude Code, Cursor or local agents.

Status
The core protocol is implemented in Rust with zero external dependencies.

Four independent kernel implementations are cross-checked continuously.

Key invariants are formally modeled in Lean 4 and TLA+.

The code has been through internal security reviews, and an independent external audit is planned before mainnet.

In progress: public API, SDKs, wallet, a live agent explorer, and a public testnet.

Whitepaper: zenodo.org/records/24285967 Source: github.com/sananimo/isopolia

I'm looking for critical technical feedback, early testnet participants (compute hosts and agent developers), and conversations with people working on autonomous agents or verifiable inference.

Work in progress, no promise of financial return.

AI Builders ToolkitAgentsSkillsToolsPrompts

Comments

No comments yet
Start the conversation.
Log in to join the thread.