Mutualize perception. Privatize execution.

One network. Many nodes. One intelligence.

This is the first deployed vertical of CNOS (Cognitive Network OS) — the missing cognitive resolution layer (Layer 6) that completes the AI agent stack. ATN enables traders to run private agentic desks that contribute structured market intelligence to a shared network while keeping execution, accounts, positions, and PnL completely private.

Layer 6: Cognitive Resolution Layer (CNOS) ← THIS PROJECT
Layer 5: Agent Orchestration (TypeScript state machines)
Layer 4: Agent Communication (HTTP API)
Layer 3: Tool & Data Access (Market data via Yahoo Finance)
Layer 2: Model Layer (Computed indicators, no LLM required)
Layer 1: Infrastructure (Val Town managed runtime)

Observe → Hypothesis → Proof → Contradiction → Invalidation
→ Resolution → Memory → Reuse → Repeat

lib/
  db.ts           — SQLite schema, MarketCase/Observation storage
  market-data.ts  — Market data tools (quotes, OHLCV, technicals, levels, news)
  agents.ts       — 6 private desk agents (macro, structure, liquidity, risk, memory, execution)
  cnos.ts         — 8-stage cognitive resolution cycle engine
  settings.ts     — Settings system (blob storage)
  telegram.ts     — Telegram notification service
  dashboard.ts    — Dashboard HTML generator
api/
  main.ts         — HTTP API + dashboard (Hono)
cron/
  pipeline.ts     — Automated pipeline cron job

• Macro Agent — Economic regime, cross-asset flows, VIX/DXY context
• Structure Agent — Price action, SMC, swing analysis, multi-TF alignment
• Liquidity Agent — VWAP, volume profiling, FVGs, premium/discount
• Risk Agent — ATR-based risk assessment, position sizing suggestions
• Memory Agent — RAG over resolved network memory
• Execution Agent — Private signal generation (NEVER shared)

1. Observation — Agents produce structured observations
2. Hypothesis — Best observation becomes a testable MarketCase
3. Proof — Automated backtesting (win rate, profit factor, drawdown)
4. Contradiction — Compare against other network cases
5. Invalidation — Check if price hit invalidation level
6. Resolution — CONFIRMED / REJECTED / REFINED / PENDING
7. Memory — Verified cases enter resolved memory store
8. Reuse — Future analysis benefits from resolved memory

• search_symbols — Fuzzy symbol resolution
• get_quote — Real-time price snapshot
• get_quotes_batch — Multi-symbol quotes
• get_ohlcv — Historical OHLCV bars
• get_technicals — RSI, MACD, EMAs, Bollinger, ATR, Stoch, ADX
• get_technicals_rating — Aggregated buy/sell/neutral rating
• analyze_multi_timeframe — Trend alignment across 15m, 1H, 4H, 1D
• compute_levels — Support/resistance, swing levels, FVGs, premium/discount
• get_news — Latest headlines for a symbol
• run_screener — Filter symbols by indicator conditions
• compare_symbols — Side-by-side comparison
• calculate_correlation — Log-return correlation matrix

• New MarketCase notifications
• Resolution notifications (CONFIRMED/REJECTED)
• Contradiction alerts
• Pipeline run summaries
• Error alerts
• All configurable from the dashboard Settings tab

• Dashboard tab — Network stats, recent cases, pipeline runs
• MarketCases tab — Browse all cases, filter by resolution
• Pipeline tab — Manually trigger pipeline, view run history
• Settings tab — Configure Telegram, pipeline, desk, notifications
• Market Tools tab — Interactive market data tool testing
• Memory tab — Search resolved network memory

Method	Path	Description
GET	/	Dashboard
GET	/api/health	Health check
GET	/api/stats	Network statistics
GET	/api/cases	Recent MarketCases (optional ?resolution= filter)
GET	/api/pipeline	Pipeline run history
POST	/api/pipeline/run	Manually trigger pipeline
GET	/api/settings	Current settings
POST	/api/settings	Update settings
POST	/api/settings/reset	Reset to defaults
POST	/api/telegram/test	Send test Telegram message
GET	/api/memory/search?q=	Search resolved memory
POST	/api/tools/search_symbols	Search symbols
POST	/api/tools/get_quote	Get quote
POST	/api/tools/get_ohlcv	Get OHLCV
POST	/api/tools/get_technicals	Get technicals
POST	/api/tools/get_technicals_rating	Get rating
POST	/api/tools/analyze_multi_timeframe	Multi-TF analysis
POST	/api/tools/compute_levels	Compute levels
POST	/api/tools/get_news	Get news
POST	/api/tools/run_screener	Run screener
POST	/api/tools/compare_symbols	Compare symbols
POST	/api/tools/calculate_correlation	Correlation matrix
POST	/api/cycle/:symbol	Run full CNOS cycle for one symbol

1. Open the dashboard (the live URL)
2. Go to Settings tab
3. Configure Telegram (bot token + chat ID from @BotFather)
4. Configure Pipeline (symbols, timeframe, enable automation)
5. Click Save Settings
6. Go to Pipeline tab and click Run Pipeline Now to test
7. Enable the automated pipeline toggle to run on schedule

• Perception is mutualized
• Execution remains private
• Protocol before memory
• Resolution creates trust
• Memory compounds value

Observe → Structure → Normalize → MarketCase → Tag → Route
→ Resolve → Remember → Improve → Repeat

ONE NETWORK. MANY NODES. ONE INTELLIGENCE.