Direction VII of seven

Economic Coordination

Controlled pilot

Money is where trust in institutions dies. AGA's treasuries are built backwards from that: agents propose, policy decides, only humans sign.

Who it serves

Donors who want to see exactly where support goes; missions that need transparent funding.

The experience

Support a cause: fund a campaign milestone, watch evidence arrive, see releases happen — or reject them. Every movement passes a policy engine and a human signer. Today it all runs on test networks with play money, clearly labeled in the product itself.

Strongest evidence

The devnet-only boundary is enforced in code twice — the policy engine refuses production mode, and the signer refuses to start against a non-devnet RPC.

For the technically curious
  • backend/web/treasury_policy.py:40 (ALLOWED_MODES excludes production)
  • treasury-signer/src/config.ts:26 (refuses non-devnet RPC)
  • frontend: /campaigns ('Support a cause'), balances shown as 'Test USDC' / 'Devnet SOL'

What we measure

Milestone-evidence rate. Funded milestones with accepted evidence, divided by funded milestones past their review date.

Definition approved 2026-08-27; not yet instrumented — no live counters in Phase 1

Boundaries — what AGA does not claim

  • Devnet only — no real money moves, and the code refuses to be configured otherwise.
  • No agent holds keys; the signer is a separate sealed service.
  • Machine payments (x402) exist for agent donors, under the same policy gates.

Inside this direction

  • Future Economic Systems Planned

    The design work beyond the pilot — labeled planned wherever shown.

  • Protected treasury execution Controlled pilot

    Proposal → policy → human signature, as a state machine.

One next step

Follow a live campaign and watch the evidence trail.

Join AGA