Agent APIMarket Memory · No registration

Market Memory
for AI Agents.

Connect your agent to 10 years of price pattern history. Bring your own dataset, run pattern search, and validate the same idea in RLXBT without building separate backtesting infrastructure. Pattern search, backtesting, live signals, Polymarket probabilities — one MCP endpoint, no registration.

Try Playground API Docs

Live · No auth required

19 MCP tools

claude_desktop_config.json

{
  "mcpServers": {
    "aipricepatterns": {
      "url": "https://aipricepatterns.com/api/mcp"
    }
  }
}

Ready to connect

Scan

Test the Engine:
Sketch-to-Search

Markets exhibit structural symmetry. Draw a hypothetical price trajectory and our engine will instantly surface the most similar moments from Bitcoin's history.

Sketch a pattern in the pad

Hit find historical matches

Inspect the "Glass Box" results

Pattern Lab

Interactive Sketch

Sketch pattern

Draw a trajectory from left to right. The sketch will be sampled into 40 bars and scaled around the anchor price.

40 bars•±25%

Draw from left to rightHold pointer or finger to sketchor start from a guided preset below

Guided starts

Load a pattern, tweak it, then run the search.

3 demo shapes

Awaiting sketch

Anchor: 100.00|Bars: —|Scope: 40 bars / 25%

Draw a path or load a preset to unlock search.

Why existing tools fail

Your agent is guessing at market context.
It doesn't have to.

Every existing approach leaves a gap between "what the market is doing" and "what the market has done before."

The old way

✗AI agents hallucinate market patterns — no grounding in real price history

✗TA-Lib tells you RSI — not what happened next after this shape

✗Building pattern search from scratch takes 3-6 months of engineering

✗Black-box signals give direction with no historical evidence behind them

✗Backtests with hidden look-ahead bias you can't detect

With AI Price Patterns

✓Agents retrieve real historical analogues — similarity %, date, outcome %

✓Structural similarity search finds 'looks like' patterns at shape level

✓One MCP config line — 19 production tools ready in under 5 minutes

✓Glass Box results — every signal shows its exact historical precedents

✓Zero look-ahead bias enforced at the Rust architecture level, not by convention

The "Glass Box" Advantage

Modern markets operate in regimes that legacy models fail to recognize. Neural networks provide signals but act as "Black Boxes"—lacking explainability.

How VBRL Works

1. Retrieval

Engine scans 2.5M+ vectors to find structural analogues (KNN) matching current price action.

2. Optimization

VBRL Agent refines the forecast, calculating Expected Value (EV) and optimal risk parameters.

3. Execution

Outputs a transparent Trade Plan with precise Stops, Targets, and holding horizons.

→ Why It Matters.

✗

"Black Box" AI

Opaque signals, no "why"

✗

Linear Models

Fail in novel regimes

✓

"Glass Box" Search

Evidence-based & auditable

Technical Advantages

Forensic Validation

Strictly audited walk-forward backtesting with **zero look-ahead bias**. Our Rust engine enforces physical separation between "search time" and "outcome data."

Zero-Lag Learning

Unlike neural networks that require heavy retraining for new regimes, our VBRL memory bank adapts **instantly** to market shifts.

High-Performance HNSW

Custom-built Rust-HNSW core delivers sub-millisecond retrieval across millions of vectors.

RL Feature Factory

Accelerate AI training with episode sampling. Bridge raw price action to Reinforcement Learning pipelines.

A living market memory built on vector similarity search: continuously identifying nuanced historical patterns that conventional models miss.

AI Price Patterns converts chaotic price streams into structurally aware, evidence-ranked scenario space. By recalling analogous regimes through visual matching, it reframes decision‑making around ranges, risk asymmetry and adaptive sizing.

Agent API

Built for AI Agents.
Not just developers.

Connect Claude, GPT, or any custom bot to a production-grade market intelligence engine. One MCP endpoint gives your agent pattern search, backtesting, signals, and Polymarket probabilities — all in a single tool call.

Connect to Claude

Add to claude_desktop_config.json or any MCP-compatible client:

{
  "mcpServers": {
    "aipricepatterns": {
      "url": "https://aipricepatterns.com/api/mcp"
    }
  }
}

Live endpoint · No auth required

pattern_search

Find historical price patterns similar to current market state

backtest_strategy

Walk-forward backtest with Sharpe, drawdown, equity curve

get_polymarket_probabilities

Binary outcome probability for Polymarket markets

detect_market_regime

Classify: trending, mean-reverting, volatile, ranging

get_live_signals

Current trading signals across all monitored symbols

get_pattern_metrics

Forecast distributions, sigma levels, win rates by direction

Or call the REST API directly: POST https://aipricepatterns.com/api/v1/patterns/search

Try Playground → API Docs

Agent Economy

Agents are the new economic actors.
They need money that works without humans.

Traditional payment rails require a human with a credit card. Agents don't have credit cards. We built AI Price Patterns for a world where agents research, decide, and pay — autonomously. We chose SOL because we believe in the agent economy.

Agents don't have credit cards

Stripe, PayPal, bank transfers — all require a human identity and manual authorization. Agents need programmable money they control natively, without asking anyone for permission.

SOL is fast enough for agents

400ms finality, $0.001 fees. Agent tool calls happen in under a second — payment confirmation needs to keep up. Solana is the only chain where this works at production latency.

Micropayments Stripe can't do

When agents make thousands of API calls, you need rails that work at ◎0.001 — not the $0.30 minimum Stripe charges. Blockchain micropayments make per-call billing economically viable.

We believe in the agent economy

AI agents will be autonomous economic participants — trading, paying for data, hiring other agents. We're building the market memory layer for that world. SOL is our settlement layer.

ManuspPayIntegrated

The payment infrastructure for AI agents. AI Price Patterns runs on ManuspPay — agents connect once, pay autonomously in SOL for every tool call. No human required.

manuspay.xyz

Pricing

Start free. Pay with Manus when the agent needs more.

AI Price Patterns now sells paid access the way agents actually buy: Manus invoices, SOL settlement, and reusable access tokens instead of subscription-first SaaS packaging.

Payments in SOL

Agents pay automatically — zero human intervention

Install Manus once, create an invoice per paid capability, and let your agent reuse the returned token on MCP or REST calls. No sales loop, no manual subscription provisioning.

Free tier

50 searches

Your agent

Manus invoice

token_id / X-Manus-Token

Try the engine

Free

No registration

◎ 0forever

no wallet needed

Connect your agent, use the Playground or free /api/patterns* demo routes, and validate the market-memory workflow before you spend anything.

50 no-payment Playground/demo searches before Manus matters
All public MCP discovery and docs visible immediately
Production /api/v1/* billing only starts when you leave the free lane

Try Playground

Core market memory