Inspect
Candles, depth, tape
Live Binance data · WebMCP-native
Real markets.
Simulated risk.
Agent-run research.
PaperQuant is a crypto research laboratory where you and Codex inspect live markets, test strategies, explain evidence, and place paper-only trades together.
No real-money pathway Checksum-traced data 22 WebMCP tools
BTC / USDT$79,604.13+1.82%
Codex is working in PaperQuantEvery action is visible and reversible
01
Market contextBTC perpetual · 1 hour
02
IndicatorsHull Suite · VWAP · RSI · ADX
03
Research runVerifying official archive checksums…
04
Explain & overlayWaiting for immutable result
Real candlesCompleted-candle policy
Paper onlyMaximum 2× simulated
01 · HUMAN + AGENTAsk in plain language.
Ask in plain language.
Watch the laboratory respond.
Codex does not click around blindly. PaperQuant exposes typed WebMCP tools connected to the same chart, account, and research state you see.
YOU ASK CODEX“Test BTC’s trend regime and show me the evidence.”
No setup scripts. No exchange keys. No copied datasets.
Configure
Market, interval, indicators
Research
Historical run with provenance
Explain
Metrics, risks, chart markers
Live observer consoleEvery safe action appears in under a second, in another tab or on another device.READ ONLY
02 · THE LABORATORY
One workspace. Six research surfaces.
Built for learning, testing, and evidence—not for sending real orders.
Charts that agents can actually operate
Search any Binance USDT market, change intervals, combine indicators, inspect completed candles, and overlay research signals.
Structured tools, not brittle clicking
Strict schemas, safety annotations, bounded outputs, cleanup, and the same controller used by every human button.
inspect_chartset_indicatorsrun_backtestshow_on_chartpaper_orderDepth and aggressor flow
Live order book imbalance, spread, recent aggregate trades, and buyer/seller-initiated volume.
Versioned algorithms and Python ideas
Run the 20-hypothesis catalog or prototype chart signals locally in an isolated browser worker.
Every result carries its data story
Source URLs, checksums, UTC normalization, missing intervals, assumptions, costs, annual results, and OOS windows.
Practice Spot and perpetuals safely
Long/short simulation, funding, reduce-only orders, live P&L, and maximum 2× simulated leverage.
03 · WHY WEBMCPThe website becomes
The website becomes
an agent-ready instrument.
PaperQuant’s tools live inside the page. When the lab is open, Codex discovers what it can inspect and change—then the browser routes approved calls back into the real product.
- No separate MCP server to install
- Tools disappear when you leave the page
- Externally sourced content marked untrusted
- Financial mutations require explicit intent
01CodexReasoning + safety
02WebMCPTyped page tools
03PaperQuantShared UI controller
04EvidenceD1 · R2 · worker
{ ok: true, data, warnings }
04 · SCIENTIFIC HONESTYMissing data is a result.
Missing data is a result.
Never an invitation to invent.
Technical correctness, data completeness, and profitability are reported separately.
Correct and reproducible
The engine and declared rules executed successfully. This never automatically means profitable.
Useful, with disclosed limits
The test ran, but missing archives, assumptions, or survivorship limits remain visible.
No credible public dataset
The hypothesis stays in the catalog with its reason. PaperQuant assigns no fabricated returns.
FAST JUDGE WORKFLOW
One prompt. A complete research loop.
This canonical study uses real, checksum-verified historical data. Once warmed, the immutable result returns almost instantly while keeping its provenance and scientific hash.
Target cached response under 2 secondsObserver is read only
PROMPT FOR CODEX
Open PaperQuant Lab and give me the read-only observer URL. Set BTCUSDT perpetual to 1h, enable Hull Suite, VWAP, RSI, and ADX, inspect the completed candles, order book, and trade tape, then run the BTC filtered EMA strategy from 2023-01-01 through 2025-12-31 with $100,000. Poll it, explain technical status, data quality, and profitability separately, then overlay the signals. Do not change my paper account.
Run it in the laboratoryPAPER
ONLY
ONLY
05 · RISK BOUNDARYReal evidence.
Real evidence.
Zero real-money pathway.
PaperQuant never requests exchange API keys, never accesses a private account, and never has a withdrawal or real-order route. Research capital and paper balances stay separate.
Public market APIs onlyExplicit intent before paper mutationsMaximum 2× simulated leverageRetry-safe idempotency keys
THE MARKET IS LIVE
Bring a hypothesis.
Leave with evidence.
Open the laboratory yourself—or ask Codex to enter through WebMCP.