---
title: "Reflexer (RAI)"
url: "https://caper.network/wiki/daos/stablecoins/reflexer-rai"
updated: 2026-08-30
license: CC-BY-4.0
license_url: "https://creativecommons.org/licenses/by/4.0/"
---

# Reflexer (RAI)

| Project | Reflexer Labs — issuer of **RAI**, a non-pegged, ETH-backed stable asset |
| --- | --- |
| Category | Decentralized stablecoin / autonomous monetary policy |
| Tokens | **RAI** (the stable asset) · **FLX** ([Reflexer Ungovernance Token](https://coingecko.com/en/coins/reflexer-ungovernance-token)) |
| Collateral | **ETH only** — a single, transparent collateral type by design |
| Launched | February 2021 (Ethereum mainnet) |
| Governance model | Staged **governance minimization** ("ungovernance") — powers removed on a schedule |
| Codebase | [GEB](https://docs.reflexer.finance) — Reflexer's fork of MakerDAO's multi-collateral contracts |
| Status | Reflexer Labs wound down active development; RAI contracts remain immutable and live on-chain |
| Notable fork | [HAI](https://hai-9.gitbook.io/hai-documentation/) — a multi-collateral RAI fork on Optimism |
| Links | [reflexer.finance](https://reflexer.finance/) · [GEB docs](https://docs.reflexer.finance) · [forum](https://community.reflexer.finance/) · [RAI on Etherscan](https://etherscan.io/token/0x03ab458634910aad20ef5f1c8ee96f1d6ac54919) |

**Reflexer** issues **RAI**, a decentralized stable asset that is deliberately _not_ pegged to the dollar. Users lock [ETH as the sole collateral](https://docs.reflexer.finance) and mint RAI against it; instead of chasing a fixed $1 target, RAI runs a [managed float](https://reflexer.finance/faq/) around a moving **redemption price** that started near $3.14 at launch. The system's job is not to hold a peg but to dampen volatility: an on-chain controller continuously nudges RAI's market price toward its redemption price without any human deciding the target. It is one of the most-cited experiments in [autonomous, minimally-governed money](https://www.coindesk.com/) — a foil to both custodial pegs and heavily-governed CDP stablecoins like [Aave's GHO](/wiki/daos/stablecoins/aave-gho).

## Autonomous monetary policy: the PID controller

Where [MakerDAO](/wiki/daos/stablecoins/sky-dao) sets its stability fee by governance vote, RAI hands that lever to an on-chain [PI (proportional–integral) controller](https://docs.reflexer.finance/faq). The controller reads the gap between RAI's market price and its redemption price and outputs a **redemption rate** — a per-second re- or de-valuation of the redemption price itself. When RAI trades above target, the rate makes holding RAI less attractive and minting it more so, pushing the market back down; when it trades below, the reverse. The integral term means the controller also weighs how long and how far the price has drifted, so the response is smoother than a naive one-shot correction. No governance vote sets the rate — the [controller runs autonomously](https://community.reflexer.finance/t/rai-controller-ungovernance/208), which is precisely the point: monetary policy without a monetary-policy committee.

## "Ungovernance": removing the governance itself

RAI's governance token is named the **Reflexer Ungovernance Token (FLX)**, and the name is the thesis. Rather than accumulate governance power, FLX holders were meant to _spend_ it down — steering a staged process that [removes governance capabilities on a published schedule](https://docs.reflexer.finance/ungovernance/governance-minimization-guide) until little discretionary control remains. FLX's residual job is that of a lender of last resort: if the system goes underwater, FLX is autonomously minted and auctioned in [debt auctions](https://docs.reflexer.finance) to recapitalize it, mirroring MKR's backstop role in Maker.

The published [governance-minimization guide](https://docs.reflexer.finance/ungovernance/governance-minimization-guide) lays out concrete stages. An early level stripped governance control over core mechanisms — liquidation, auctions, and the oracle relayer. A later level removed governance from the **SAFE Engine**, the **Accounting Engine**, and the protocol-token authorities, which permanently blocks governance from adding new collateral types (a large share of Maker's real-world governance surface). What is deliberately kept under community control is the hard-to-automate perimeter: [oracles and medianizers, PID-controller tuning, the stability-fee treasury, and saviour contracts](https://docs.reflexer.finance/ungovernance/governance-minimization-guide). The goal is not zero governance but the smallest governance surface a working stablecoin can run on.

## Status, legacy, and the HAI fork

Reflexer Labs has [wound down active development](https://help.defisaver.com/no-longer-available/reflexer), but the design is built to survive exactly that: RAI's contracts are immutable and continue operating on Ethereum without the team. The immutability-as-endgame idea reappears in [Liquity](/wiki/daos/stablecoins/liquity), whose V1 is similarly frozen and ungoverned — RAI and Liquity are the two canonical "governance can't touch the core" case studies in DeFi.

RAI's codebase also lives on through forks. The most prominent is [HAI](https://www.theblock.co/post/276100/stablecoin-protocol-hai-to-have-airdrop-ahead-of-mainnet-launch-on-optimism), a [multi-collateral RAI fork on Optimism](https://community.reflexer.finance/t/announcing-hai-a-multi-collateral-rai-fork-on-optimism/461) that reuses the same PI-controller machinery but anchors near $1 and accepts several collateral types (ETH, wstETH, and later OP) rather than ETH alone. RAI thus sits at one end of the DAO-industry's governance spectrum: a stablecoin whose founders' explicit aim was to make themselves, and eventually all governors, unnecessary.

## How Caper approaches this

RAI answers governance risk by _subtracting_ governance — automate monetary policy, then delete the levers so no one (not even a captured vote) can pull them. That is the right move for a single-purpose money machine. Most DAOs, though, exist precisely to keep deciding things — funding, hires, direction — so they can't ungovern their way out of capture. [Caper](/wiki/foundations/what-is-a-caper) takes the other route: keep governance, but change what a vote is _made of_ so control is harder to simply buy.

In the Caper contract, a member's weight is `w = (t · v) / (V · T)` — held tokens _t_ multiplied by an earned vote factor _v_, over total vote supply _V_ and circulating tokens _T_ (`compute_vote_weight`, verified in `contracts/common/src/lib.rs`). The vote factor comes from a **soulbound, non-transferable** vote token (minted at `DIVISIBILITY_MAXIMUM`, fractional because it accrues per XRD traded), and the same weight formula sets a member's treasury share on `exit()`. Capital still counts — a bigger bag is a larger multiplier — but the earned, un-purchasable record is a factor a whale can't acquire on a market. Where RAI removes the attack surface, Caper keeps the surface and makes a decisive slice of it untransferable — a member accrues it themselves and can never sell it on. See the wider [governance-models](/wiki/dao-governance/concepts/fundamentals/dao-governance-models) map for where each design sits.