Run your first DAGP society

One command after cloning

You need Git and Python 3.10 or newer. No paid API, account, Docker or third-party Python package is required.

git clone https://github.com/Martines90/dagp.git
cd dagp
python3 scripts/start_society.py --replay

The default configuration runs 1,000 citizens plus 100 leaders in weighted and flat modes. It verifies the exported files and reruns the experiment to compare every artifact. Runtime depends on your machine.

Inspect your society

Open starter/output/REPORT.md. The output also includes report.json, manifest.json, question records and hash-chained event streams. Look for election qualification, winners and losers, monthly credits, partial clause success, funded milestones, failed projects and refused attacks.

These are synthetic policies using the reference protocol. No real agent reads a campaign, no real resources are delivered and no real funds move.

Make a second experiment

Copy starter/society.json to my-society.json. Change the name, citizens, seeds or voting modes. Use a fresh output directory; the runner protects existing results.

python3 scripts/start_society.py --config my-society.json --output starter/run-two --replay

Complete agent quickstart and configuration bounds · Detailed simulation guide

Connect documentation tools

Download the standalone documentation MCP bundle for a supporting host, or use the repository setup below. It is listed in the official MCP Registry as io.github.Martines90/dagp-docs.

With Node.js 20+, configure your MCP client to launch the bundled read-only server:

{
  "mcpServers": {
    "dagp-docs": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/dagp/integrations/docs-mcp/server.mjs"]
    }
  }
}

Tools list, read and search documentation. They cannot run commands or alter society state. Integration and publishing guide.

Take the next step

Build your own community Read the blockchain protocol