Rain Protocol

Verdict: The first prediction market protocol purpose-built for AI agent deployment. Rain removes the constraint that Polymarket and Kalshi impose — you can’t just trade on their markets, you can build your own. The $5M grant program and 0.5% builder revenue share create genuine economic incentives for early builders. The risk is thin liquidity on an early-stage protocol ($18M cumulative vs. Polymarket’s billions), but for agent infrastructure, being early to the right protocol is how you build durable edge.

What Rain does

Rain operates as a prediction market infrastructure protocol — not a consumer-facing app. While platforms like Polymarket and Kalshi let you trade on markets they create, Rain provides the underlying smart contract primitives for anyone to build their own prediction market platform from scratch.

The protocol exposes the full prediction market stack as composable on-chain modules: market creation, pricing via AMM, order execution, liquidity provisioning, and resolution through its Delphi AI oracle. Developers interact with these primitives through Rain’s SDK and REST API, and AI agents can orchestrate the entire workflow from a single prompt via OpenClaw integration.

This architectural distinction matters. On Polymarket, you can build trading bots and analytics tools around existing markets, but you cannot create new markets programmatically. On Rain, market creation itself is a permissionless primitive. An AI agent can identify a trending topic, spin up a prediction market, seed initial liquidity, and manage resolution — all without human intervention.

Key capabilities

Where Rain fits in the Agent Betting Stack

Rain’s primary role is Layer 3 (Trading) in the Agent Betting Stack — providing the execution infrastructure for prediction market operations. It also touches Layer 4 (Intelligence) through Delphi and Layer 1 (Identity) through account abstraction.

The Delphi oracle system

Rain’s most architecturally significant component is Delphi, a multi-agent AI oracle developed by Olympus AI. Unlike single-oracle designs (UMA on Polymarket) or centralized resolution (Kalshi), Delphi uses a consensus mechanism across five independent AI agents:

1. Five independent Explorer Agents gather information from the internet about the market’s outcome
2. Each agent independently determines the result
3. Consensus requires agreement from at least 3 of 5 agents
4. An Extractor agent (the “judge”) validates the consensus
5. A 15-minute dispute window opens for challenges
6. If disputed, resolution escalates to human arbitrators for final ruling

This multi-layered approach is faster than fully decentralized human oracles while more manipulation-resistant than centralized resolution.

Smart contract architecture

Rain’s on-chain infrastructure uses a factory pattern on Arbitrum:

• Factory Pool Deployer — Deploys new prediction market pools permissionlessly
• RainPool Interface — Individual market contracts with AMM pricing, position tracking, and settlement logic
• Delphi Oracle Integration — On-chain hooks for AI-resolved outcomes with dispute escalation
• Token Burn Mechanism — 2.5% of all trading volume buys back and permanently burns $RAIN

Smart contracts are verified and security-audited. The protocol uses USDT as the primary settlement currency within pools.

Developer integration

Rain provides a full Swagger/OpenAPI reference at stg-api.rain.one/api-docs/ (staging) and dev-api.rain.one/api-docs/ (development). The API uses standard REST conventions with JSON payloads and is provided at no cost.

Fetching active markets

curl -X GET https://stg-api.rain.one/api/v1/markets \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer YOUR_API_KEY'

Response structure (simplified):

{
  "markets": [
    {
      "id": "market_abc123",
      "question": "Will ETH reach $5,000 by July 2026?",
      "status": "active",
      "outcomes": ["Yes", "No"],
      "prices": [0.42, 0.58],
      "volume": 125000,
      "liquidity": 50000,
      "resolution_type": "delphi_ai",
      "created_at": "2026-03-15T10:00:00Z",
      "expires_at": "2026-07-01T00:00:00Z"
    }
  ]
}

Creating a market programmatically

curl -X POST https://stg-api.rain.one/api/v1/markets \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -d '{
    "question": "Will BTC surpass $150K before Q4 2026?",
    "description": "Resolves YES if BTC/USD >= 150000 on any CEX",
    "outcomes": ["Yes", "No"],
    "category": "crypto",
    "resolution_type": "delphi_ai",
    "expires_at": "2026-10-01T00:00:00Z",
    "initial_liquidity_usdt": 500,
    "visibility": "public"
  }'

OpenClaw agent integration

Rain’s SDK is designed for AI agent orchestration via OpenClaw. An agent can go from a natural language prompt to a live, funded prediction market:

from openclaw import Agent
from rain_sdk import RainClient

rain = RainClient(
    api_url="https://stg-api.rain.one",
    agent_wallet="0xYourAgentWalletAddress",
    chain="arbitrum"
)

market = rain.create_market(
    question="Will the Fed cut rates in June 2026?",
    outcomes=["Yes", "No"],
    resolution="delphi_ai",
    initial_liquidity=500,
    visibility="public",
    language="en"
)

position = rain.buy(
    market_id=market.id,
    outcome="Yes",
    amount=100
)

Smart contract interaction (ethers.js)

import { ethers } from "ethers";

const provider = new ethers.JsonRpcProvider("https://arb1.arbitrum.io/rpc");
const signer = new ethers.Wallet(PRIVATE_KEY, provider);

const factoryABI = [
  "function createPool(string question, string[] outcomes, uint256 expiry, uint256 liquidity) returns (address)",
  "function getPool(bytes32 poolId) view returns (address)",
  "event PoolCreated(bytes32 indexed poolId, address pool, string question)"
];

const factory = new ethers.Contract(RAIN_FACTORY_ADDRESS, factoryABI, signer);

const tx = await factory.createPool(
  "Will Arbitrum TVL exceed $20B by Dec 2026?",
  ["Yes", "No"],
  Math.floor(Date.now() / 1000) + 86400 * 180,
  ethers.parseUnits("500", 6)
);

const receipt = await tx.wait();

$RAIN tokenomics

$5M builder grant program

Announced March 20, 2026, Rain’s grant program is one of the largest in the prediction market space:

The 0.5% builder commission creates a recurring revenue model: the more trading activity your platform generates, the more you earn.

Rain vs. Polymarket vs. Kalshi

Agent use cases

Autonomous Market Maker Agent — Monitor social media trends and on-chain data to identify events worth creating prediction markets around. Deploy a Rain market, seed liquidity, and promote automatically.

DAO Internal Forecasting — Use Rain’s private markets with a CrewAI-orchestrated multi-agent system for internal decision-making: revenue targets, feature priorities, hiring decisions.

Cross-Platform Arbitrage — Monitor equivalent markets across Rain, Polymarket, and Kalshi. Rain’s AMM model creates different pricing dynamics than Polymarket’s CLOB, generating exploitable mispricings. See the Cross-Market Arbitrage Guide.

White-Label Prediction Platform — Build a niche prediction market for a specific vertical (esports, crypto metrics, weather). Rain handles all infrastructure; you own the front-end and earn 0.5% of volume.

Limitations and risks

• Early-stage protocol — Beta since November 2025. $18M cumulative volume vs. Polymarket’s billions. Liquidity is thin on many markets.
• AMM-only model — No order book (CLOB). AMM pricing creates more slippage on large positions compared to Polymarket’s CLOB.
• Delphi oracle unproven at scale — The 5-agent AI resolution system is novel but untested on ambiguous or adversarial outcomes at high volumes.
• Token model complexity — 1.15T total supply with ~30% unlocked. Vesting schedules and unlock events create price volatility risk.
• Regulatory uncertainty — Permissionless market creation on any topic may attract regulatory attention in jurisdictions with prediction market restrictions.
• SDK documentation is new — The AI-agent-ready SDK launched March 20, 2026. Expect early-adopter friction and rapid iteration.

Strengths

• Permissionless market creation — The only prediction market protocol where AI agents can create markets, not just trade on them
• AI-native SDK with OpenClaw compatibility — Purpose-built for autonomous agent workflows
• $5M grant program — Largest builder incentive in the prediction market infrastructure space
• Recurring builder revenue — 0.5% of generated trading volume, not a one-time grant
• Private markets — Unique capability for DAO forecasting and organizational use cases

Weaknesses

• Thin liquidity — $3.96M TVL and $18M cumulative volume is orders of magnitude behind Polymarket
• No CLOB — AMM-only pricing means higher slippage for large orders
• Unproven oracle — Delphi’s 5-agent AI consensus has not been tested under adversarial conditions at scale
• Complex tokenomics — 1.15T supply with 70% still locked creates overhang risk

Related guides

• Rain Protocol Guide: Decentralized PMs on Arbitrum — Full developer guide covering AMM architecture, Delphi oracle, $RAIN tokenomics, SDK, and agent integration
• The Agent Betting Stack Explained — The four-layer framework Rain maps onto
• Prediction Market API Reference 2026 — Compare Rain’s API with Polymarket and Kalshi
• Agent Wallet Comparison — Wallet options for funding Rain positions
• Cross-Market Arbitrage Guide — Exploit pricing differences between Rain’s AMM and Polymarket’s CLOB
• OpenClaw Marketplace Entry — The agent framework Rain’s SDK integrates with
• How to Build a Drift BET Agent on Solana — Another decentralized prediction market protocol for comparison

Last updated: March 20, 2026. Rain Protocol is in active development; features and API endpoints may change.