meridian: autonomous agents for RWA and long-tail liquidity

Constant-product AMMs assume continuous liquidity. Real-world RWAs do not. meridian is our open research line on agent-driven quoting, inventory risk, and verifiable pricing inputs.

meridian
Open-source research artefact · MIT

The problem with one-size AMMs for RWAs

Pool-based AMMs work when flow is dense and price discovery is on-chain. For many RWAs — real estate slices, invoices, regional commodities — liquidity is patchy, oracles are authoritative, and spread must reflect jurisdiction and counterparty risk. A passive LP curve is the wrong abstraction.

What meridian is

meridian is a Python framework for autonomous market-making agents: they observe oracles and inventory, adjust spreads, simulate scenarios in a backtesting engine, and can be extended with geography- and asset-class-specific policies.

  • Oracle-driven reference pricing and sanity checks
  • Adaptive spreads and inventory skew
  • Backtesting against historical or synthetic paths
  • Hooks for compliance and venue-specific rules

Research angle

We treat meridian as a living artefact: every strategy is falsifiable against data, and every deployment assumption is documented. Partners can fork policies without forking the safety rails.

Docs & code

Full reference documentation lives on docs.kcolbchain.com/meridian. Source and issues: GitHub.