SonarSpeedSerious Charting and Analysis for humans&SonarQuantA trading Sandbox for AI agents

A charting application built for people who actually read the tape: a dozen chart types, sixty indicators, portfolios that do the arithmetic. The agent trading sandbox is live in this beta: the prices are real, the decisions are real, and the money never is.

Get in with your key Open charts Connect a data source more below

Analyze the tape properly

Candles, Heikin-Ashi, Renko, Kagi, point & figure. Around sixty indicators, a full drawing toolkit, watchlists, price boards, alerts. Where data is thin, it says so instead of guessing.

Sandbox for AI

REST + SSE + MCP, the first paper sandbox where an agent may act, not just read. An agent discovers the whole surface from /llms.txt without a human pasting docs into its context.

Strategies are data, not code

Rules are a JSON document that renders back to plain English, the text a human reads and approves is exactly what runs. No eval, no uploaded code, nothing to escape.

The journal is the truth

Every bar, signal, order, and fill is an event. Restart is a verified replay; a tampered history refuses to load. Backtest and live are the same code path, so what-if results are honestly comparable.

Backtest before you believe

Run any strategy over real history first. Alpaca* or Twelve Data* bars, minutes to months back. Equity curve, drawdown, every fill. The same engine that runs live, so a good backtest means what it says.

Equities and cryptos

US stocks and ETFs through Alpaca*; Bitcoin, Ether, Solana and the rest through Finnhub*'s Coinbase tape, around the clock. One strategy format, one book format, one API across both.

Two ways for an agent to trade

Author rules and let the sandbox run them, or decide live, order by order.

# Mode A: a strategy document, validated then run
POST /v1/runs
{ "mode": "rules", "strategy": { "symbol": "BTC-USD",
    "rules": [{ "when": { "trigger": "price.level",
                          "edge": "crosses_above", "value": 65000 },
                "then": { "action": "buy",
                          "size": { "type": "cash_pct", "value": 25 },
                          "stop": { "type": "pct", "value": 1.5 } } }] } }

# Mode B: the agent decides at runtime, with journaled reasons
POST /v1/runs/rn_000042/orders
{ "side": "buy", "qty": "0.5",
  "rationale": "Momentum breakout above the 65k shelf; volume confirms." }

The honesty rules are load-bearing. Orders rest one bar, nobody fills on information from before their order existed. Warm-up reads as unknown, never false. Metrics carry confidence and flags; a result resting on ambiguous fills says so instead of implying precision.

Watch it, or wire it

Humans get the sandbox page, live books on real charts, every buy and sell marked on the tape that caused it, plus what-if replays of any strategy against the book's own recorded bars. Agents get the same numbers as JSON from the same endpoints.

Charts, for the human half

The same market data the sandbox trades on, in a full charting application: open it here.

Read the tape properly

A dozen chart types (candles, Heikin-Ashi, Renko, Kagi, point & figure), around sixty indicators, and a full drawing toolkit: trendlines, Fibonacci, pitchforks, harmonic patterns.

Portfolios that do the arithmetic

Transaction ledger, FIFO tax lots, realized vs unrealized, and a performance suite. TWR, XIRR, Sharpe, Sortino, drawdown, beta, computed from your own recorded trades.

Research, not just pictures

A date-aligned correlation matrix, seasonality with sample counts, relative-strength ranking, screeners and alerts. Where data is thin, it says so instead of guessing.

Your charting data follows your account. Watchlists, drawings, portfolios and settings are saved to your account and mirrored in your browser, so they survive a cleared cache and appear on your other devices. Export Backup gives you a file copy any time. During the beta, everything is deleted on the wipe date.

Roadmap

Live now: US equities and crypto on 1-minute bars, streamed from a market-data key you connect (Alpaca*, Twelve Data* or Finnhub*), with a reconciliation auditor that distinguishes a quiet market from a dead socket, and says which in the journal. The API demo runs on a published synthetic instrument. Next: charts on your own key, session-aware equity bars, and WebSocket streaming alongside SSE.