Integration Guide

Give Claude Market Memory

Connect Claude to 10 years of Bitcoin price pattern history in under 5 minutes. One config line — Claude gains 19 MCP tools for pattern search, backtesting, live signals, and Polymarket probabilities.

Setup in 3 steps

No registration. No API key. Paste and go.

1

Open Claude Desktop config

Find claude_desktop_config.json in ~/Library/Application Support/Claude/ on macOS.

2

Paste one config block

Add the aipricepatterns entry under "mcpServers". That's the entire setup.

3

Restart Claude

Restart Claude Desktop. All 19 tools load automatically — no auth, no keys.

claude_desktop_config.json
{
  "mcpServers": {
    "aipricepatterns": {
      "url": "https://aipricepatterns.com/api/mcp"
    }
  }
}
19 Tools Available

What Claude can do after connecting

Ask Claude anything about price patterns. It will call the right tool automatically.

"Find patterns similar to the last 30 candles"

Searches 2.5M+ vectors for structurally similar historical moments.

"What happened after this pattern historically?"

Returns outcome distribution from all matching historical episodes.

"Backtest a breakout strategy on BTC"

Runs walk-forward backtests with zero look-ahead bias.

"What's the Polymarket probability for BTC above 70k?"

Pattern-driven probability estimates for active prediction markets.

"Show live signals for the current market"

Real-time pattern match scores across multiple timeframes.

"Detect the current market regime"

Classifies trending, ranging, or reversal regimes from recent structure.

Glass Box — not a black box

Core differentiator

Every answer Claude gives comes with the exact historical episodes behind it. Not just "the pattern looks bullish" — but which exact dates, similarity %, and what happened in the next N candles. Claude can show its work.

Every pattern match includes the original date, similarity score, and n-candle outcome
No hallucinated signals — every result is grounded in the actual price history database
Zero look-ahead bias enforced at the Rust architecture level, not by convention
Claude can explain exactly why it's bullish or bearish in a given context
Autonomous billing

Claude pays automatically — no human intervention

Install the Manus binary on your server once. It runs as an MCP middleware — Claude calls tools as usual, Manus handles the SOL billing in the background.

Claude
MCP tool call
Manus middleware
SOL deducted
Tool executes
1

Download Manus binary

One binary, installs on your server. No dependencies, no Docker, no separate service.

2

Configure once

Point Manus at your AI Price Patterns endpoint and fund it with SOL. Takes under 2 minutes.

3

Agents pay themselves

Claude connects as usual. Manus intercepts each tool call, bills in SOL, passes through. No human action needed.

Example Claude conversation

After connecting, this is the kind of analysis Claude can deliver instantly.

Claude Desktop
U

You

BTC just formed a tight consolidation after a 15% rally. Find similar historical patterns and tell me what happened next.

C

Claude

I'll search the pattern database for structurally similar moments. Found 12 matching episodes (similarity > 82%): • 2021-02-14 — 84% similarity → +18.3% over next 7 days • 2020-10-29 — 82% similarity → +24.1% over next 7 days • 2019-06-22 — 83% similarity → -4.2% over next 7 days Outcome distribution (n=12): Bullish (>5%): 8/12 cases — 67% Neutral (±5%): 3/12 cases — 25% Bearish (<-5%): 1/12 cases — 8% Median outcome: +11.4% · Avg drawdown before upside: 2.1% Current regime: trending-up (confidence 0.78) Historical base rate strongly favors continuation. The one bearish case occurred during a macro risk-off event in March 2020.

Ready to give Claude market memory?

Free tier, no registration. Works with Claude Desktop and any MCP-compatible client.