Invite-Only Beta

Every AI trade.
Recorded. Checked. Measured.

QuantNova is the control plane between your AI agent and its trading workflow. One MCP connection creates an append-only decision record, runs deterministic risk checks, and measures every call against what the market did next.

QuantNova Canvas Pipeline
Live 24/7 Append-only audit 3 markets
1
MCP Connection
3
Markets · US A HK
11
Risk Guards
T+1/7
Outcome Scoring
24/7
Always Recording

Not a black box. Fully transparent.

Your agent can already trade. These are the three things it's missing.

Every decision recorded

Your Claude logs every buy, sell, hold, and watch — with the full reasoning and the signals it relied on. Append-only ledger; nothing can be quietly rewritten.

One record across three markets

US, A-share, and Hong Kong decisions use a consistent audit format while retaining market-native data and risk context.

Scored against reality

Every decision is evaluated at T+1 and T+7 against real market data. Hit rate, guard flags, behavior red flags — you finally know if your agent is actually good.

Every decision, fully visible.

Click any node to drill down. From macro view to individual trade — nothing is hidden.

Trading Pipeline
Three Markets
Governed Learning
10-step trading pipeline

10 transparent steps

From signal fusion ranking to 6-role AI debate to 11 guard checks to order execution. Every step produces data you can inspect. Search any symbol to trace its full journey through the pipeline.

Gating → Debate → Guard → Execute → Verify
US macro A-share macro

Same pipeline, different intelligence

Switch to US: Fear & Greed, Fed Rate, Treasury Yield. Switch to A-share: northbound flow sentiment, margin trading balance. The entire tree restructures per market — not just different labels.

US: Finnhub · A: AKShare + Tencent · HK: Tencent
Self-evolving feedback loop

Learning with evidence and controls

IC and decision accuracy produce proposed weight updates with versioned evidence, monitoring, and a clear rollback path. Automation stays observable instead of silently rewriting history.

IC → Fusion weights · Accuracy → Debate weights

While you sleep, it's working.

A typical day — automated where safe, reviewable at every step.

Pre-open

Morning Briefing

Market outlook, key themes, and what matters for your portfolio today

Intraday

Anomaly Detection

Significant moves caught in real time — alerts come to you

Auto

6-Role AI Debate

6 analysts assess independently, contradictions flagged and resolved

Close

End-of-Day Report

What moved, why it moved, and what to watch tomorrow

Weekly

Self-Improvement

Weights adjusted, strategies refined — the system evolves based on real performance

The longer it runs, the more evidence you have

Track outcomes, compare versions, and improve with a record you can inspect.

Always Evolving Signal Generate Track Outcome Learn From results Improve Get sharper

Connect your agent in 5 minutes.

Works with Claude Code, Claude Desktop, and any MCP-compatible agent.

01

Create a key

Request access, log in, and open the Agent Audit page. One click creates your qn_ API key — shown once, revocable anytime.

02

Add the MCP server

One command connects your Claude to the Recorder:

claude mcp add --transport http quantnova https://quantnova.app/mcp --header "Authorization: Bearer qn_..."
03

Trade. It's all on the record.

Ask your agent to analyze or trade anything. Every decision lands in your ledger with its reasoning — and gets scored at T+1 and T+7 automatically.

Your AI can trade now.
Make sure you can trust it.

See why your AI trades what it trades — not just what.

Have an invite code? Activation is immediate.

Not ready to connect an agent? Join the product update list.