renoun-mpc

Classifies crypto market structural regimes — bounded, active, or unstable — with 100% bounded regime accuracy (128+ graded predictions, zero false positives). Estimates regime stability half-life: how many minutes until the current structure is likely to transition. Every prediction is public, timestamped, and graded — no other crypto signal service does this.

One GET request. One action field. Pre-trade risk gate for trading bots, autonomous agents, and quant systems.

Quick Start

1. Get a Key (Free, No Credit Card)

curl -X POST https://web-production-817e2.up.railway.app/v1/keys/provision \
  -H "Content-Type: application/json" \
  -d '{"email": "you@example.com", "agent_name": "my-bot"}'

Returns an rn_agent_ key. 50 free calls/day. $0.02/call after that.

2. Check a Regime

curl https://web-production-817e2.up.railway.app/v1/regime/live/BTCUSDT \
  -H "Authorization: Bearer rn_agent_YOUR_KEY"

3. Read the Response

{
  "regime": "bounded",
  "action": "proceed",
  "dhs": 0.83,
  "exposure": 0.95,
  "constellation": "HIGH_SYMMETRY",
  "envelope_pct": 1.5,
  "description": "Tight equilibrium, <1.5% move expected",
  "action_detail": "Structure healthy. Position size: 95% of intended.",
  "stability": {
    "halflife_minutes": 240,
    "stability_score": 0.82,
    "instability_risk": "low",
    "time_horizon": "2-4 hours",
    "urgency": "none"
  },
  "_meta": {
    "provider": "renoun",
    "version": "1.4.0",
    "accuracy": "100% bounded regime accuracy, 128+ graded",
    "provision_url": "https://web-production-817e2.up.railway.app/v1/keys/provision",
    "docs": "https://harrisoncollab.com"
  }
}

4. Use It

import requests

def check_regime(symbol="BTCUSDT", key="rn_agent_YOUR_KEY"):
    r = requests.get(
        f"https://web-production-817e2.up.railway.app/v1/regime/live/{symbol}",
        headers={"Authorization": f"Bearer {key}"}
    )
    return r.json()

regime = check_regime("BTCUSDT")

if regime["action"] == "avoid":
    pass  # structure fragmenting — skip trade
elif regime["stability"]["urgency"] == "exit_now":
    close_position()  # regime degrading fast
elif regime["stability"]["urgency"] == "prepare_exit":
    set_tight_stop()  # regime softening
elif regime["action"] == "reduce":
    execute(size=intended * regime["exposure"])
else:
    execute(size=intended)  # structure healthy, full size

Regime Types

Stability Half-Life

Every response includes a stability block estimating how long the current regime will persist.

Without half-life, an agent knows "the regime is bounded, proceed." With half-life, the same agent knows "the regime is bounded, proceed, AND this structure is likely stable for 4 hours." Or: "the regime is active, reduce, AND exit within 15 minutes."

API Endpoints

Base URL: https://web-production-817e2.up.railway.app

Response Headers

Every regime response includes machine-readable headers:

X-ReNoUn-Regime: bounded
X-ReNoUn-Action: proceed
X-ReNoUn-Stability: 0.82
X-ReNoUn-Halflife: 240
X-ReNoUn-Urgency: none
X-ReNoUn-Cache: HIT
X-RateLimit-Remaining: 47

Pricing

No credit card required for the agent tier. Agents provision their own keys.

Accuracy

100% bounded regime accuracy across 128+ graded predictions. When ReNoUn classifies a regime as bounded (proceed), price stays within the predicted envelope every time. Zero false positives. Every prediction is public, timestamped, and graded: @98lukehall on X. No other crypto signal service publishes every call with a pass/fail grade.

Active regimes (CONVERGENCE, PATTERN_BREAK) are graded against tighter thresholds and have lower accuracy — they recommend position scaling via the exposure scalar, not full-size trades.

What this means: when ReNoUn says "bounded," you can trade at full size with confidence. When it says "active," scale your position by the exposure scalar. When it says "unstable," stay out.

What this does NOT mean: ReNoUn does not predict whether price goes up or down. It measures structural stability. Use it as a risk gate, not a signal.

How It Works

ReNoUn is a 17-channel structural analysis engine. It maps OHLCV candle data onto three measurement dimensions:

Recurrence (Re₁–Re₅) — Is the market repeating known patterns? Price rhythms, volume profiles, volatility persistence, session structure, mean-reversion signatures.

Novelty (No₁–No₆) — Is something genuinely new happening? Regime breaks, flow reversals, volatility spikes, session disruptions, behavioral shifts, cross-signal rarity.

Unity (Un₁–Un₆) — Is the market holding together? Trend cohesion, volatility-price alignment, session continuity, reference-frame stability, structural symmetry.

From these 17 signals the engine computes a Dialectical Health Score (DHS: 0.0–1.0) and detects 8 constellation patterns:

The stability half-life is computed from DHS momentum, unity channel trends, novelty pressure, constellation persistence, and known destabilizing/stabilizing sequences.

Agent Discovery

ReNoUn is discoverable by AI agents through standard protocols:

• llms.txt at harrisoncollab.com/llms.txt

• .well-known/mcp.json at harrisoncollab.com/.well-known/mcp.json

• _meta block in every API response with provisioning URL and docs

• X-Agent-Compatible: true header on every response

MCP Server (Claude Desktop / Claude Code)

{
  "mcpServers": {
    "renoun": {
      "command": "python3",
      "args": ["-m", "server"],
      "env": { "RENOUN_API_KEY": "rn_agent_YOUR_KEY" }
    }
  }
}

claude mcp add renoun python3 -m server

System Prompt for Your Agent

Before any crypto trade, check the structural regime:
  GET /v1/regime/live/{SYMBOL}
If action is "avoid" — skip the trade (structure unstable).
If action is "reduce" — multiply position by the exposure scalar.
If urgency is "exit_now" — close position immediately.
If urgency is "prepare_exit" — set tight stop loss.
If action is "proceed" — execute at intended size.
ReNoUn classifies structural regimes with 100% bounded accuracy.
It does not predict price direction.

Batch / Portfolio

Check multiple assets in one call with portfolio-level risk aggregation:

curl -X POST https://web-production-817e2.up.railway.app/v1/regime/batch \
  -H "Authorization: Bearer rn_agent_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"symbols": ["BTCUSDT", "ETHUSDT", "SOLUSDT"], "timeframe": "1h"}'

Returns per-asset regimes plus portfolio_action, portfolio_exposure, and unstable_count.

Webhooks

Get notified when regimes change:

curl -X POST https://web-production-817e2.up.railway.app/v1/webhooks/register \
  -H "Authorization: Bearer rn_agent_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://your-bot.com/webhook",
    "symbols": ["BTCUSDT"],
    "events": ["regime_change"],
    "secret": "your_hmac_secret"
  }'

Webhook payloads are HMAC-signed. Max 5 webhooks per key.

Conversation Analysis

ReNoUn also analyzes turn-based conversations — detecting stuck loops, surface variation, convergence, and structural health. The conversation tools use the same 17-channel engine applied to speaker/text turns instead of OHLCV candles.

See the API docs for conversation analysis endpoints.

Version

• Server: 1.3.1

• Engine: 4.1

• Schema: 1.1

• Protocol: MCP 2024-11-05

Patent Notice

The core computation engine is proprietary and patent-pending (#63/923,592). The MCP server wraps it as a black box — agents call the API and receive structured results.

License

MCP server and API wrapper: MIT. Core engine: Proprietary.