Copilot

AI-powered one-call investment verdicts, portfolio analysis, and thesis tracking — the highest-value synthesis product for FIRM's investment research suite.

What's inside#

Copilot distills 24+ underlying data layers (ML predictions, voter decomposition, smart-money flows, analyst consensus, options activity, insider trades, and macro regime) into single, structured investment verdicts. One call returns:

Single-ticker analysis — blended score (-1 to +1), decile rank, 90% conformal interval, 12-voter breakdown, position-sizing recommendation, and a self-contained narrative the LLM can quote verbatim or rewrite.
Portfolio scoring — aggregate factor exposure across holdings, identify tilts (momentum, value, quality, sector concentration), flag concentration risks, and suggest top-3 names to trim or add.
Portfolio-aware verdicts — when evaluating a candidate ticker against existing holdings, adjusts position size for correlated holdings and sector overweight, detects diversification opportunities, and returns a decision (BUY / BUY-WITH-CAUTION / HOLD / PASS).
Track record — out-of-sample IC (information coefficient), top-decile Sharpe, conformal coverage, and DSR-adjusted significance tests from the model's live monthly backtest.

Data sources: real-time market data, 12 quant voters (ML ensemble, macro context, fundamental value, technical, regime HMM, sentiment, options flow, insider transactions, analyst revisions, anomaly scan, and two proprietary advanced technical models), 30-day average daily volume (tradeability filter), earnings-blackout annotations, and Tengu's proprietary 36-month realized-return correlation mesh for portfolio-aware risk assessment.

Access#

10 credits/call. Unlocked by Enterprise plan or the $199/mo Copilot add-on (available on Pro and below).

Endpoints#

Method	Path	Description
GET	`/api/v1/copilot/score/{ticker}`	Full per-ticker analysis: blended score, decile, conformal interval, 12-voter breakdown, regime context, suggested position size, narrative.
POST	`/api/v1/copilot/portfolio`	Score a list of user-held tickers; aggregate factor tilts; identify top-3 names to trim/add.
GET	`/api/v3/copilot/ticker_full/{ticker}`	Omnibus single-stock aggregation: transparency (ML drivers, prediction, calibration, voter IC drift, coverage) + smart-money (13F changes, institutional ownership, insider trades, options flow, dark pool, max pain, GEX) — 12 layers fanned in parallel, 1 HTTP request.

Examples#

Bash#

Get a single-ticker verdict:

Shell

1curl -H "Authorization: Bearer $TENGU_API_KEY" \2  "https://firm.tengu.co/api/v1/copilot/score/AAPL"

Post a portfolio for scoring:

Shell

1curl -X POST -H "Authorization: Bearer $TENGU_API_KEY" \2  -H "Content-Type: application/json" \3  -d '{4    "holdings": [5      {"ticker": "AAPL", "weight_pct": 0.15},6      {"ticker": "MSFT", "weight_pct": 0.12},7      {"ticker": "NVDA", "weight_pct": 0.10}8    ]9  }' \10  "https://firm.tengu.co/api/v1/copilot/portfolio"

Response envelope (all copilot endpoints):

JSON

1{2  "ok": true,3  "timestamp": "2026-07-05T14:32:15.123456Z",4  "ticker": "AAPL",5  "as_of_ts": "2026-07-05T09:30:00Z",6  "knowledge_ts": "2026-07-04T23:59:59Z",7  "blended_score": 0.234,8  "decile": 7,9  "decile_label": "above-median (decile 7)",10  "rank": 892,11  "conviction": 0.72,12  "voter_coverage": 0.92,13  "voters_supporting": 8,14  "voters_against": 2,15  "expected_return_pct": 3.21,16  "interval_lo_pct": 0.5,17  "interval_hi_pct": 6.1,18  "interval_half_width_pct": 2.8,19  "interval_method": "conformal",20  "tier": "large",21  "regime": "normal",22  "suggested_position_pct": 0.045,23  "voter_breakdown": [24    {25      "voter": "ml_ensemble",26      "score": 0.312,27      "weight": 0.25,28      "contribution": 0.078,29      "interpretation": "LightGBM v2 forecasts a positive 1m return"30    },31    {32      "voter": "macro_context",33      "score": 0.156,34      "weight": 0.15,35      "contribution": 0.023,36      "interpretation": "macro environment supports risk-on positioning at the 1m horizon"37    }38  ],39  "narrative": "AAPL: bullish signal (blended score +0.234, above-median (decile 7)). 8 voters supporting, 2 against; top contributors: ml_ensemble, macro_context, fundamental. 90% conformal interval for next 1m return: +0.5% to +6.1%. In a normal-regime market, the model's edge is weakest. Suggested position size: 4.5% NAV (half-Kelly, capped 8%).",40  "model_track_record": {41    "oos_ic_1m": 0.044,42    "oos_sharpe_top_decile": 1.18,43    "conformal_coverage_realized": 0.895,44    "n_months_oos": 18,45    "track_record_stale": false46  },47  "adv_dollars": 8500000.0,48  "in_earnings_blackout": false49}

Python#

Get a ticker verdict and use the narrative:

Python

1import os, requests2 3r = requests.post(4    "https://firm.tengu.co/api/v1/copilot/portfolio",5    headers={"Authorization": f"Bearer {os.environ['TENGU_API_KEY']}"},6    json={7        "holdings": [8            {"ticker": "AAPL", "weight_pct": 0.15},9            {"ticker": "MSFT", "weight_pct": 0.12},10        ]11    },12    timeout=30,13)14r.raise_for_status()15envelope = r.json()16print(f"Timestamp: {envelope['timestamp']}")17print(f"Portfolio score: {envelope['portfolio_score']:+.3f}")18print(f"Weighted decile: {envelope['weighted_decile']:.1f}/10")19print(f"Coverage: {envelope['coverage_pct']:.0%} of holdings in prediction universe")20print(f"Factor tilts: {envelope['factor_tilts']}")21print(f"Narrative: {envelope['narrative']}")

API reference — complete endpoint catalog with per-route credit costs and plan access.
Pricing & credits — plan tier details, add-on pricing, and credit consumption.
Quant Signals — underlying 12-voter decomposition, ML predictions, and conformal intervals (Pro+).