BTCBoard

Bitcoin multi-agent research board for evidence-backed BTC market decisions.

BTCBoard is not a live trading bot. It is a paper-trade-only research workbench that recreates a professional BTC trading desk:

Question
  -> Market data adapters
  -> TA / Sentiment / OnChain / Macro agents
  -> Evidence Board
  -> Strategy debate
  -> Risk-Agent veto
  -> Paper Trade Plan
  -> Backtest / Reflect
  -> Bitcoin-native access layer

Current Status

BTCBoard has completed the full MVP roadmap through the Bitcoin-native layer.

• Phase 1: Core Agents complete.
• Phase 2: EvidenceItem and Evidence Board complete.
• Phase 3: structured signal system complete.
• Phase 4: trading team frontend board complete.
• Phase 5: signal logging, paper-trade backtest, and Reflect-Agent complete.
• Phase 6: MCP / L402 / Nostr-NWC Bitcoin-native access complete.
• Final integration layer: report persistence, history API, Markdown/JSON export, and MCP history/export tools complete.

Core Agents

• TA-Agent: OHLCV, RSI, MACD, EMA, volume, volatility, support/resistance.
• Sentiment-Agent: news sentiment, ETF narrative, regulatory event detection.
• OnChain-Agent: mempool fee and activity adapter, exchange-flow placeholder.
• Macro/Event-Agent: event risk and macro calendar context.
• Strategy-Agent: bull case, bear case, base case, final bias.
• Risk-Agent: veto-capable risk gate that can pass, reduce, wait, or block.
• Reflect-Agent: paper-trade review and agent weight suggestions.

Output Shape

BTCBoard produces a full research report with:

• signal
• bias
• confidence
• entry_condition
• invalidation
• risk_level
• position_hint
• review_trigger
• evidence_board
• paper_trade_suggestion
• backtest
• bitcoin_native
• export_artifacts

The system intentionally supports wait_confirmation and no_trade; it is designed to reject weak trade ideas.

Backend

Run from:

cd D:\ir\BTCBoard-main\BTCBoard-main

Create a venv if you need to install backend dependencies:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r backend\requirements.txt

Start the API:

$env:PYTHONPATH="backend"
python -m uvicorn app.main:app --reload --host 127.0.0.1 --port 8000

Frontend

cd D:\ir\BTCBoard-main\BTCBoard-main\frontend
npm install
npm run dev

The frontend is a trading workbench, not a landing page:

• top market bar,
• left research input,
• center agent cards and strategy debate,
• right Evidence Board,
• bottom Paper Trade Plan, Reflect Log, Bitcoin-native Access, Persistence & Export.

API

• GET /health
• POST /api/research
• POST /api/agent/research
• GET /api/history
• GET /api/export/markdown
• GET /api/export/markdown/{report_id}
• GET /api/export/json
• GET /api/export/json/{report_id}
• GET /.well-known/agent.json
• GET /.well-known/mcp.json
• GET /api/bitcoin-native
• GET /api/l402/offer
• POST /mcp

MCP Tools

• btcboard.run_research
• btcboard.get_status
• btcboard.get_history
• btcboard.export_report

Bitcoin-Native Layer

BTCBoard exposes a safe Bitcoin-native integration layer:

• MCP manifest for other agents.
• Optional L402 gate.
• Nostr identity placeholder.
• NWC read-only wallet observation placeholder.
• Hard safety posture: paper_trade_only, live execution disabled.

Useful environment variables:

BTCBOARD_L402_ENABLED=true
BTCBOARD_L402_REQUIRED=false
BTCBOARD_L402_PRICE_SATS=21
BTCBOARD_L402_PREIMAGE_SECRET=
BTCBOARD_NOSTR_ENABLED=false
BTCBOARD_NOSTR_NPUB=
BTCBOARD_NOSTR_RELAYS=wss://relay.damus.io,wss://nos.lol
BTCBOARD_NWC_ENABLED=false
BTCBOARD_NWC_CONNECTION=

Persistence

Local JSONL data is stored under:

• backend/data/signals.jsonl
• backend/data/reports.jsonl

Each research run creates a report record and, when a signal exists, a paper-trade signal record.

Validation

Verified locally:

python -m compileall backend\app

Manual smoke checks passed for:

• council run,
• report persistence,
• history summary,
• Markdown export.

Frontend dependencies are not vendored; run npm install before building or starting the UI.