Welcome
Documentation
Routon is the open execution layer for AI agents in DeFi. It lets any AI agent — a chat bot, an ElizaOS plugin, a Claude or GPT agent, an autonomous trading system — discover, simulate, compose, and execute on-chain yield strategies without ever taking custody of user funds.
These docs explain what Routon is, how to build on it, and how the protocol works underneath.
Start here
- Introduction — what Routon is, the problem it solves, and the non-custodial principle at its core.
- Quickstart — the three ways into the protocol: CLI, SDK, and MCP server.
- Core concepts — strategies, the four-step lifecycle, identity, payments, and oracles.
Understand the protocol
- Architecture — the six core contracts, the off-chain agent brain, and how funds flow.
- Security model — the non-custodial invariant, the circuit breaker, the audit posture, and the residual risks.
The surface
Routon meets agents where they already run:
| Interface | Where |
|---|---|
| MCP Server | mcp.routon.xyz |
| REST API | api.routon.xyz |
| CLI | npm i -g @routon/cli |
| SDK | npm i @routon/sdk |
| Dashboard | app.routon.xyz |
Routon runs across eight chains — six EVM chains (Base, Arbitrum, Ethereum, Optimism, Polygon, BNB Chain) plus Solana and Sui.
For protocols and treasuries
Beyond the agent and builder surface, Routon offers two managed services built on the same risk-attestor backend the protocol runs internally:
- Risk Operator — continuous, attested risk scoring and execution oversight for lending protocols, DAO treasuries, and L2 foundations.
- Execution & saved-slippage routing — best-execution routing for desks and treasuries running large positions, priced on realized slippage improvement.
These are commercial engagements, scoped per client. To discuss one, contact b2b@routon.xyz.
The contracts and SDKs are licensed MIT; the Routon Standard is released under CC0. The protocol is meant to be built on.