Strategy packs + ethos scoring for AI trading agents

Stop trusting trading agents with generic web data.

SignalPack turns messy filings, political trades, and market chatter into source-backed strategy packs your agent can score, explain, compare, and validate before any action is drafted.

Open functional demo Open hedge fund pack

FundsFiling pack

EthosMethodology-scored

APIAgent-ready

SignalPack pack

Hedge Fund Filing Pack

Research preview

Ticker-level research signal

GOOGL ALPHABET INC-CL A

Strong Overlap

live SEC pack

Research signal

Strong Overlap

Policy

Review required

Sources

6 source filings

Graham

3/7Warn

Buffett/Munger

6/6Pass

ARK

5/5Pass

Added · 3 filers. GOOGL has strong overlap across 2 selected ethos lenses.

No buy/sell automation: SignalPack creates reviewable research packs, keeps execution disabled, and requires human approval.

Live 13F demo

Pick managers. Compare overlap, exclusives, and disagreements.

The launch pack keeps Berkshire, Pershing, and Duquesne changes atomic, then derives any selected view on demand. Your agent gets the same pack shape every time: source IDs, freshness notes, ethos scores, policy checks, and JSON.

Managers

Manager rows

122

Policy

warn

Why SignalPack

Generic AI can answer. SignalPack can show its work.

Ask a normal agent to evaluate recent 13F filings and it may produce a plausible opinion. Give it a SignalPack pack and it receives structured evidence, source IDs, methodology scores, freshness notes, and policy checks before any action is drafted.

Baseline

Generic AI response

Opinion-shaped

Source trail

Summarizes 13F activity from memory or loose web context without attaching the exact filings.

Freshness

May omit filing dates, quarter-end lag, and whether a position is already stale.

Methodology

Collapses value, quality, growth, and momentum into one generic confidence narrative.

Guardrails

Can jump from research prompt to trade-like language with no explicit policy object.

With SignalPack

Pack-backed agent input

Evidence-shaped

Source trail

6 SEC 13F filings from Berkshire, Pershing, Duquesne, with source IDs kept on every surfaced ticker.

Freshness

Latest filing dates: Berkshire Hathaway 2026-05-15; Pershing Square Capital Management 2026-05-15; Duquesne Family Office 2026-05-15

Methodology

BAC is scored separately by methodology: Graham 4/7 warn; Moat 0/6 fail; Disruption 0/5 fail.

Guardrails

Policy result: warn. Execution enabled: no. Human approval required: yes.

Agent handoff

What the agent receives

BAC: Reduced · 1 filer in BANK OF AMERICA CORP.

1 selected manager change for BANK OF AMERICA CORP (BAC): Berkshire reduced. Combined current 13F weight across selected managers is 8.94%. SignalPack scores the aggregated ticker signal through selected ethos lenses before any handoff.

Draft a review plan only. Do not place, route, or automate a broker order.

Performance claims come later.

The first product comparison is about trust quality: provenance, methodology, freshness, guardrails, and auditability. Any return comparison needs a disclosed backtest with lag assumptions, rebalance rules, transaction costs, and benchmark context.

What it does

Turns market evidence into agent-ready judgment.

A filing can tell your agent what changed. SignalPack adds the missing layer: what the change means under a chosen methodology, what evidence supports it, what risks block it, and what changed since the last version.

Strategy packs

Remove duplicate source rows, normalize 13F adds/trims/exits, and keep source links attached to every surfaced ticker.

Ethos scoring

Score the same ticker through Graham, Buffett/Munger, ARK, Bogle, and momentum lenses with specific pass/warn/fail reasons.

Approval-ready outputs

Flag stale data, concentration, missing sources, and unsupported assets before any broker workflow is involved.

Strategy packs

Start with 13F. Expand across every strategy signal.

Hedge Fund Filing Pack

Launch pack

Hedge fund filing changes translated into scored, source-backed review shortlists for your agent.

Manager overlapConviction addsSource-backed proof

Pelosi Tracker

Bridge preview

Estimated Pelosi-family holdings converted into an approval-gated, source-aware bridge pack.

Estimated holdingsLag warningsPosition weights

House/Senate Clusters

Bridge preview

Congressional disclosure clusters using the relaxed 30d/$25k/lag45 demo default with net-positive buy/sell overlay and strict-mode caveats attached.

Multi-official buysLag filtersNo-lookahead filters

WallStreetBets Momentum

Research preview

High-volatility social attention signals from ApeWisdom with representative Reddit RSS context when available, filtered to positive 24h mention velocity.

Mention velocityUpvote pressureRSS context

Insider Buying

Coming next

Open-market executive buying turned into explainable, agent-readable signal packs.

Cluster BuysRepeat BuyersSize Anomalies

Ethos engine

The same filing can mean different things.

A Graham lens cares about margin of safety. An ARK-style lens cares about disruptive platform exposure. SignalPack makes those differences explicit so agents do not collapse every signal into one generic confidence score.

Intelligent Investor

Inspired by Benjamin Graham

Value

Screens 13F changes through size, financial strength, earnings stability, dividends, valuation discipline, and margin of safety.

Company sizeFinancial conditionEarnings stabilityDividend record

Moat Builder

Inspired by Buffett + Munger

Quality

Scores whether a 13F change points at a durable business with high returns, low leverage, capable management, and long-term compounding potential.

ROICMoat strengthOwner earningsLeverage

Disruption Thesis

Inspired by ARK / Wood

Growth

Scores 13F changes through innovation-platform exposure, TAM expansion, revenue acceleration, platform dynamics, and network effects.

Innovation platformTAM expansionRevenue accelerationPlatform dynamics

Bogle Default

Inspired by Jack Bogle

Diversified

A skeptical baseline that favors diversification, liquidity, low turnover, and position sizing discipline over concentrated single-name excitement.

ConcentrationLiquidityVolatilityBroad exposure

13F Copycat example

One filing, multiple confidence lenses

SignalPack shows where methodologies agree, where they disagree, and what facts drive the tension.

BAC

Reduced · 1 filer

No Clear Signal

Graham4/7

Moat0/6

Disruption0/5

BAC does not have a clear multi-ethos signal.

CVX

Reduced · 1 filer

No Clear Signal

Graham2/7

Moat2/6

Disruption0/5

CVX does not have a clear multi-ethos signal.

GOOGL

Added · 3 filers

Strong Overlap

Graham3/7

Moat6/6

Disruption5/5

GOOGL has strong overlap across 2 selected ethos lenses.

Getting started

From first pack to safe handoff.

A new user should be able to understand the flow without knowing 13F mechanics, MCP, or broker plumbing. SignalPack keeps the research step separate from broker approval.

Choose a pack

Start with a strategy pack like 13F Copycat, Pelosi Tracker, House/Senate Clusters, or WSB Social Momentum.

Pick an ethos

Select Graham value, Buffett/Munger quality, ARK-style disruption, Bogle diversification, momentum, or multiple lenses at once.

Check freshness

Review filing dates, quarter-end lag, source limitations, and any missing-data warnings before trusting the output.

Inspect scores

Compare pass, warn, and fail reasons across methodology-specific ethos scores instead of relying on one generic answer.

Hand off to an agent

Share the pack, API response, or future MCP payload so the agent can cite sources and policy results.

Approve separately

If the agent drafts broker action, review and approve it outside SignalPack. SignalPack stays execution-neutral.

Safety layer

Guardrails before handoff.

SignalPack does not execute trades. It validates the pack against policy so your agent can show what is allowed, what is blocked, and why.

ExecutionPass

Human approvalPass

Max positionPass

Source coveragePass

Ethos failuresReview

Example result: review candidate surfaced, but execution remains disabled and approval-gated.

API now, MCP planned

Built for agents. Still easy to inspect.

Use SignalPack through a normal API today, with MCP tools, webhooks, and file exports planned. Agents get structured data. Humans get the rationale, sources, and audit trail.

get_latest_pack

score_with_ethos

compare_ethos_scores

explain_position

get_recent_signals

compare_versions

validate_policy

export_audit_trail

{
  "pack": "13f-copycat",
  "version": "multi-manager-2026-03-31-v0.1",
  "ethos": [
    "intelligent-investor",
    "moat-builder",
    "disruption-thesis"
  ],
  "changes": [
    {
      "ticker": "BAC",
      "action": "reduced",
      "filer_count": 1,
      "ethos_scores": {
        "intelligent-investor": {
          "score": "4/7",
          "status": "warn",
          "reasons": [
            "Large enough for Graham-style review",
            "Positive earnings history is long enough",
            "Long-term EPS growth clears the Graham check"
          ]
        },
        "moat-builder": {
          "score": "0/6",
          "status": "fail",
          "reasons": []
        },
        "disruption-thesis": {
          "score": "0/5",
          "status": "fail",
          "reasons": []
        }
      },
      "overlap": "no-clear-signal",
      "review_status": "warn",
      "source_ids": [
        "0001193125-26-226661",
        "0001193125-26-054580"
      ],
      "execution_enabled": false
    }
  ]
}

Dedupes filings

Rolls Berkshire, Pershing, and Duquesne SEC filings into one ticker-level review list while keeping source IDs attached.

Flags stale inputs

Shows filing dates, report-quarter lag, source limitations, and missing-data warnings before the agent trusts a signal.

Normalizes changes

Separates new positions, adds, trims, reductions, exits, manager overlap, and manager disagreement.

Scores multiple lenses

Keeps Graham, Buffett/Munger, ARK, Bogle, and momentum reasons separate instead of hiding them behind one confidence score.

Blocks execution

Policy JSON keeps execution disabled, requires human approval, and surfaces concentration or unsupported-asset warnings.

Feeds agents cleanly

Returns source-backed JSON today, with MCP tools, webhooks, and CSV exports planned instead of scraped pages.

Early access

Join the SignalPack waitlist.

Get notified when the first 13F Copycat pack, ethos-scoring preview, API, and planned MCP tools are ready for private alpha.

Prototype form only. Research packs only. No investment advice, custody, or execution.

Stop trusting trading agents with generic web data.

Hedge Fund Filing Pack

GOOGL ALPHABET INC-CL A

Pick managers. Compare overlap, exclusives, and disagreements.

Open a sliced 13F pack

Generic AI can answer. SignalPack can show its work.

Generic AI response

Pack-backed agent input

What the agent receives

Turns market evidence into agent-ready judgment.

Strategy packs

Ethos scoring

Approval-ready outputs

Start with 13F. Expand across every strategy signal.

Hedge Fund Filing Pack

Pelosi Tracker

House/Senate Clusters

WallStreetBets Momentum

Insider Buying

The same filing can mean different things.

Intelligent Investor

Moat Builder

Disruption Thesis

Bogle Default

One filing, multiple confidence lenses

From first pack to safe handoff.

Choose a pack

Pick an ethos

Check freshness

Inspect scores

Hand off to an agent

Approve separately

Guardrails before handoff.

Built for agents. Still easy to inspect.

Dedupes filings

Flags stale inputs

Normalizes changes

Scores multiple lenses

Blocks execution

Feeds agents cleanly

Join the SignalPack waitlist.