SEC Insider Trading Tracker

Behavioral insider intelligence from SEC Form 4 filings.

SEC Insider Trading Tracker is a real-time SEC Form 4 API for insider-trading monitoring, behavioral signal classification, and event-driven alert workflows.

Most insider-trading tools surface transactions. This actor classifies behavioural intent.

It suppresses compensation noise, detects coordinated insider activity, tracks issuer regime shifts, and produces deterministic decision-ready signals for automation and institutional workflows.

Built on raw SEC Form 4 XML filings. No LLMs in the decision layer. Fully auditable, replayable, routing-safe.

SEC Insider Trading Tracker is a behavioral insider-trading signal platform — a developer-first insider trading API and monitoring platform with deterministic signal classification, compensation-noise suppression, cluster detection, and workflow-ready escalation primitives. Designed for hedge funds, compliance surveillance teams, quant researchers, corporate IR teams, and AI-agent workflows.

Real-time Form 4 monitoring with webhook-ready deterministic escalation outputs. Pair with Apify Schedules for live drift detection, regime-shift alerts, and event-driven downstream workflows — every new Form 4 filing emits a signalVerdict.classification + decisionReadiness.state gate that automation can branch on without parsing prose.

A CEO conviction buy and a tax-withholding sale should not enter a workflow with equal weight. Most insider tools still treat them the same.

---

Why insider data fails most investors

Most insider-trading feeds overwhelm users with low-quality activity:

• tax-withholding sales,
• routine options exercises,
• scheduled 10b5-1 dispositions,
• compensation grants,
• low-materiality director sales.

The problem is not access to Form 4 filings. The problem is separating behavioural signal from reporting noise.

This actor is engineered to solve that distinction deterministically.

Raw Form 4 filings are data. Behavioural interpretation is the edge.

---

What this replaces

Insider workflows usually live across three places — none designed for the question "which of these actually matters?":

Tool category	Strength	Where it fails
Bloomberg / FactSet / CapIQ	Coverage, terminal UX, institutional trust	Filter-and-sort UX. No noise suppression. No behavioural interpretation. Analyst-driven synthesis.
OpenInsider / RSS / web trackers	Free, comprehensive, public	No prioritization. No signal-quality. No cluster intelligence. No workflow primitives.
Internal hedge-fund pipelines	Bespoke conviction scoring + routing	Fragmented. Analyst-specific. Non-portable. Usually not deterministically auditable.

This actor productizes the institutional reasoning hedge funds usually build internally — deterministic conviction scoring, behavioural archetype classification, signal-decay modelling, decision-readiness gates, contradiction detection — and ships it as a stable JSON contract that downstream automation branches on.

It is deterministic decision infrastructure with honest nullability: every classification is reproducible; every external dependency is named with its sibling actor.

---

What a high-quality signal actually looks like

A grade-A CEO open-market buy emits:

signalVerdict.classification: high-conviction-bullish
decisionReadiness.state:      thesis-ready
severity:                     critical
escalationRecommendation:     immediate-review
signalGenome:                 [ceo, open-market-conviction-buy, critical-materiality, rare, low-noise, founder-conviction]

A coordinated 4-officer disposal emits:

signalVerdict.classification: distribution-risk
decisionReadiness.state:      thesis-ready
severity:                     critical
cluster.pattern:              officer-cluster-sale
regimeState.current:          accelerating-distribution-regime
signalGenome:                 [senior-officer, strategic-sell, critical-materiality, rare, officer-cluster-confirmed, low-noise, distributor]

A routine tax-withholding sale emits:

signalVerdict.classification: compensation-noise
decisionReadiness.state:      noise
severity:                     informational
noiseClassification.suppressRecommended: true
signalGenome:                 [insider, sell, low-materiality, low-noise]

Same Form 4 source. Three radically different routing decisions. The classifier is deterministic — the same input always produces the same verdict.

---

SignalGenome — compressed behavioural recognition

SignalGenome is the most extreme expression of trust-weighted compression in the schema — every record carries a compact behavioural fingerprint:

["ceo", "open-market-conviction-buy", "critical-materiality", "rare", "officer-cluster-confirmed", "low-noise"]

The same fields that drive routing also drive recognition. SignalGenome compresses insider role + conviction quality + rarity + cluster participation + materiality + noise level + behavioural deviation into a single human-scannable structure optimized for:

• Slack alert subjects
• dashboard cells
• PM review queues
• agent tool-call summaries
• compliance triage

Reading the genome takes ~1 second. Reading the equivalent 6 separate fields takes ~10. This is the perception primitive most insider tools lack.

The schema is optimized for operational decision compression, not just data extraction.

---

What downstream systems branch on

Most downstream automation only needs 5 fields:

Field	Use
signalVerdict.classification	9-value enum — direction + strength + noise tag
decisionReadiness.state	Production-routing gate (only thesis-ready drops into auto-action)
severity	6-tier institutional taxonomy (informational → critical)
escalationRecommendation.level	Routing tier (immediate-review / weekly-review / archive / ignore)
signalGenome	Token-scannable fingerprint for human + agent recognition

These five fields compress hundreds of underlying behavioural and regime signals into deterministic routing primitives suitable for Slack escalation, portfolio review queues, compliance workflows, Dify agents, Zapier filters, and PM dashboards. This is the trust-weighted compression the actor's downstream consumers actually act on.

The full record carries 100+ fields for analysts who want depth. The 5 fields above are what production routing actually consumes.

This actor converts insider activity into deterministic routing primitives.

---

Who this is for

Audience	Primary use
Hedge fund research	cluster-buy + officer-disposal intelligence, watchlist drift detection
Compliance / surveillance teams	replayable insider-activity audit, regime-shift escalation
Quant researchers	structured behavioural features for alpha models
Corporate IR teams	governance-deterioration detection on portfolio companies
AI-agent builders	deterministic routing primitives, signalVerdict-driven automation
Activist / short-side researchers	leadership-consensus breakdown, executive-vs-board divergence
Retail investors	rare officer cluster buys + distress-pattern detection (with high-noise suppression)

Each audience reads the same record differently. The institutionalNarrative block ships PM / risk-committee / compliance / short-seller views per record so the same data routes correctly to each desk.

Built for PM review queues, analyst escalation, portfolio surveillance, real-time insider-trading alert pipelines (Slack / PagerDuty / webhooks / Zapier / Make / n8n), event-driven AI-agent tool calls, and live Form 4 monitoring via Apify Schedules.

---

Detect

• Strategic insider buying — rare officer buys, founder-confidence patterns, broad accumulation
• Executive disposal waves — multi-officer cluster sales, accelerated dispositions
• Coordinated selling — same-week participation, hierarchy spread, timing cohesion
• Regime shifts — accumulation → liquidation, executive-confidence-recovery, governance-deterioration
• Rare insider behaviour — first-time CEO buys in years, deviation from an insider's profile, extreme materiality
• Noise vs signal — automatic suppression of tax-withholding, scheduled 10b5-1, compensation grants

---

The difference between noise and signal

The rare insider signals that matter are usually buried inside compensation noise.

Most insider-trade tools surface every Form 4 transaction equally. This actor classifies them by behavioural quality — a tax-withholding sale and a CEO conviction buy show up differently:

Filing	Generic tools	This actor
Tax withholding (code F)	"Insider sold $500k"	noiseClassification.noiseType: "tax-withholding", signalVerdict.classification: "compensation-noise", suppressRecommended: true
CEO open-market buy	"CEO purchased shares"	behaviorProfile.profileType: "founder-confidence-buy", signalVerdict.classification: "high-conviction-bullish", severity: "critical"
4-officer sell within 14 days	"4 insiders sold"	cluster.pattern: "officer-cluster-sale", regimeState.current: "accelerating-distribution-regime", signalVerdict.classification: "distribution-risk"
Routine options exercise	"Insider exercised options"	intentClassification.primaryIntent: "compensation-vesting", signalQuality.tier: "background-noise", suppressRecommended: true
Director cluster buy after price drop	"3 directors bought"	cluster.coordinationSignals.timingCohesion: "tight", signalVerdict.classification: "bullish", regimeState.current: "accumulation-regime"

---

Decision-output stack (what each block answers)

The output is engineered for trust-weighted compression — the same data, stratified for different reading speeds:

Reading speed	What you read	Question answered
5 seconds	signalVerdict.classification + severity + decisionReadiness.state	"Should I look at this right now?"
30 seconds	decisionCards[] + signalCards[] + signalGenome + oneLine + headline	"What's the headline finding for the run?"
5 minutes	Full record — behaviorProfile + intentClassification + regimeState + confidenceDrivers + contradictionSignals + narrativeDelta	"Why does the verdict say what it says?"
Reviewer	decisionSnapshot + runManifest + inputsHash	"Is this decision reproducible / auditable?"

---

Deterministic decision infrastructure

This actor is engineered as auditable institutional intelligence, not black-box AI finance magic. The trust architecture is the core product — every classification, score, and prediction is deterministic, replayable, and proactively honest about its own limits.

Unlike opaque AI scoring systems, this actor uses deterministic behavioural classification over SEC filing patterns. The actor is designed for institutional workflows where explainability and replayability matter more than opaque scoring. The system surfaces uncertainty explicitly instead of masking it behind aggregate scores.

Trust surfaces

Surface	What it does
confidenceCalibration	Surfaces overconfidence proactively — cold-start records carry overconfidenceRisk: "high" even when raw decision.confidence reads above 0.8
modelIntegrity	Per-record + run-level integrity (classificationIntegrity / dataCompleteness / externalDependencyRisk)
manipulationResistance	Detects staggered-disposal + wash-pattern + fragmentation. gamingResistance: "high" records are clean for production routing
contradictionSignals	Internal-classifier inconsistency detector — records with high-severity contradictions can never reach thesis-ready
decisionReadiness.state	Production-routing gate — only thesis-ready records drop into auto-action queues
signalAttribution (reserved)	Reserved structure for v5 outcome-validation joins via finnhub-stock-data sibling — buyers see the v5 contract baked in today
decisionSnapshot	Every record carries eventId + inputsHash + rulesApplied + profileSnapshot. Same input + same profile = same eventId across runs

Multi-perspective reading

The same record reads differently depending on who's consuming it. The institutionalNarrative block ships four perspectives per transaction — portfolioManagerView (long-thesis support / pressure), riskCommitteeView (concentration limits + risk flags), complianceView (routine vs flag), shortSellerView (short-thesis support / contradiction). A grade-A insider buy reads as "long-thesis support" to a PM AND "short-thesis contradiction — re-examine entry" to a short-seller. Same data, four deterministic-template frames.

Anticipatory layer

forwardEscalationRisk predicts the next-14-day trajectory: probabilityOfEscalation14d (0-1) + expectedNextState + likelyDrivers[] + confidence. The prediction is heuristic + deterministic — same inputs always produce the same probability, every driver code is on the record, and high-severity contradictions proactively suppress the prediction (the actor doesn't predict when it doesn't trust its own classification).

Honest nullability

For fields requiring external data this actor doesn't have — market cap, earnings calendars, multi-year reputation history — the field ships with null and a note pointing at the sibling actor that delivers it (finnhub-stock-data, sec-edgar-filing-analyzer). No fabrication. Buyers see what's promised AND what's reserved for v5+ in the same schema.

Honest nullability is the trust contract: every promised field that can't be honestly populated today is reserved with a sibling pointer instead of fabricated.

Confidence is capped when behavioural history is insufficient. Signals with internal contradictions cannot escalate automatically.

Production-grade contracts

• Stable enum tokens — every classification field uses a documented enum that's additive across minor versions
• Composable JSON-only output — six record types, one stable contract
• Routing-safe — designed for Slack, PagerDuty, Zapier, agent tool calls. The actor produces routing primitives; the customer's automation consumes them
• Watchlist mode — opt-in watchlistName + named KV store enables cross-run drift, regime-shift detection, cold-start → mature progression. Restricted-permission tokens probe-once and gracefully degrade
• Low false-positive design — suppression engine + contradiction detection + cold-start caps + noise-aware scoring conspire to keep high-conviction-bullish rare. When it fires, it's worth reading

---

Architecture — five-layer deterministic decision infrastructure

Layer	Per-record question	Example fields
Extraction	What does the SEC filing say?	sharesTraded, pricePerShare, totalValue, transactionType
Intelligence	What kind of signal is this?	direction, materiality, unusualness, insiderRole, transactionSignals[]
Operational	What's happening across the issuer + cohort?	clusterMembership, issuerState, behaviorFingerprint, triggersFired[]
Decision-infrastructure	Should a human look at this RIGHT NOW?	decision, recommendedAction, reviewSla, escalationRecommendation, priorityQueue
Temporal	What changed since last run?	temporalSignals, delta, stateTransition, regimeShift, narrativeDelta, runDelta

Every transaction emits recordType: "transaction". Every detected cluster emits recordType: "cluster". Every issuer emits recordType: "issuer-summary". Every run ends with one recordType: "batch-summary" carrying portfolio state, priority queue, watchlist analytics, and run manifest.

---

What makes a strong signal?

Most insider-trade tools surface every transaction equally. A $5M CEO buy and a $50k tax-withholding sale both show up as "insider activity." That's noise.

This actor classifies transactions by behavioural quality, not just dollar value. The strongest signals carry multiple converging factors:

Signal pattern	Why it matters
Rare open-market officer buy — CEO/CFO/COO buying with personal money, direct ownership, after months/years of no buys	Senior officers rarely buy on the open market — when they do, it usually correlates with management confidence in upcoming fundamentals. The actor flags these as behaviorProfile.profileType = "founder-confidence-buy" or "conviction-buyer".
Multi-officer cluster buy — 2+ distinct senior officers buying same direction within a 30-day window	Coordinated officer accumulation is one of the highest-quality bullish signals in insider data. The actor flags clusters with coordinationSignals.executiveConcentration = "high" and pattern = "officer-cluster-buy".
Strategic distribution by senior officer — large disposal by CEO/CFO with critical materiality, post-vesting period	Senior-officer disposals at scale can signal sentiment shift or planned liquidity — the materiality + intent classifier separates routine 10b5-1 from discretionary distribution.
Coordinated reduction across hierarchy — multiple officers + directors selling within days, broad hierarchy spread	The coordinationSignals.hierarchySpread[] field surfaces when the cluster crosses CEO/CFO/Director boundaries — a stronger signal than any individual sale.
Behavioural deviation — insider activity that breaks their established pattern (first buy in years, sudden cluster after silence, accelerated disposal vs scheduled cadence)	The behaviorProfile.behaviorShift and unusualness.factors[] arrays make deviation explicit. Watchlist mode accumulates per-insider history for stronger deviation detection across runs.
High materiality + high unusualness + low noise — composite signal where dollar size, position impact, behavioural rarity, and signal-quality scoring all converge	The signalQuality.tier = "institutional-grade" (score ≥ 80) is the headline filter for high-conviction analysts and quants.

Weak signals (auto-suppressed):

• Tax-withholding (transaction code F) — automatic disposal to cover vesting tax, not discretionary
• Grant/award (code A) — equity compensation, not an open-market signal
• Routine options exercise (code M/C) — typically scheduled vesting activity
• Small disposals by non-senior insiders — likely portfolio rebalancing or 10b5-1 cadence
• Low-materiality grants from compensation cycles — predictable comp activity, no signal

The noiseClassification.suppressRecommended = true boolean is the production-routing primitive — downstream automation can branch WHERE noiseClassification.suppressRecommended = false to filter out compensation noise without silently dropping the records.

Behavioural archetypes

Every transaction carries a behaviorProfile.profileType enum. Same insider × same dollar value, different archetype = different signal. The classifier is deterministic: pattern detection over the insider's run-level activity (and watchlist-level history when watchlist mode is on).

Bullish archetypes

• strategic-accumulator — repeated direct buys across multiple transactions, all direct ownership
• conviction-buyer — single high-materiality officer open-market buy with no offsetting sells
• opportunistic-buyer — material officer buys mixed with prior dispositions
• crisis-accumulator — accumulation pattern during high-pressure issuer state (requires watchlist context)
• founder-confidence-buy — CEO direct open-market buy with critical materiality

Neutral / compensation archetypes

• compensation-cycler — grants + immediate withholding, dominant compensation activity
• routine-options-exerciser — repeat options exercises with no offsetting buys
• scheduled-seller — repeated similar-size sales consistent with a 10b5-1 plan
• liquidity-manager — mixed buy/sell with no dominant pattern

Bearish archetypes

• strategic-distributor — sell-dominant senior-officer activity with critical materiality
• accelerated-disposal — multiple senior-officer sales within a short window
• governance-exit-pattern — senior-officer dispositions correlated with departure indicators (full detection requires watchlist context)
• pre-event-deleveraging — sells preceding catalyst events (full detection requires earnings-calendar integration via finnhub-stock-data sibling)

---

Why use SEC Insider Trading Tracker

Corporate insiders -- officers, directors, and shareholders who own 10% or more of a company's stock -- are legally required to report their transactions to the SEC within two business days via Form 4 filings. These disclosures are among the most closely watched signals on Wall Street because insiders have privileged knowledge of their own company's operations, pipeline, and financial health. When insiders buy with their own money, it often reflects genuine confidence in the company's future. When they sell large blocks, it can signal concern or a simple need for liquidity.

Manually browsing the SEC EDGAR website to find and interpret these filings is tedious and time-consuming. Each Form 4 is filed as a raw XML document that requires careful parsing to extract the meaningful data -- who traded, what they traded, how many shares, at what price, and how much they still hold. This actor automates the entire pipeline from search to structured output, saving hours of manual work per research session.

Whether you are an individual investor screening for insider buying signals, a compliance team monitoring portfolio companies, a financial journalist investigating unusual activity, or a quant researcher building trading models, this actor gives you programmatic access to the same insider trading data used by institutional investors and hedge funds -- at a fraction of the cost of commercial data providers.

---

Key features

• Insider behavioural fingerprinting — Every transaction carries a behaviorProfile block with one of 14 archetypes (strategic-accumulator, founder-confidence-buy, conviction-buyer, compensation-cycler, accelerated-disposal, strategic-distributor, …) + confidence + signalReliability + activity-pattern array. Pure deterministic classification — no LLMs, no external APIs.
• Transaction intent classification — intentClassification.primaryIntent separates discretionary-conviction-buy from tax-obligation from scheduled-10b5-1 from liquidity-stress-sale (12 stable intents). Lets users filter WHERE intentClassification.primaryIntent IN ('discretionary-conviction-buy', 'opportunistic-accumulation') to skip compensation noise.
• Noise suppression engine — noiseClassification.suppressRecommended: boolean flags compensation cycles, tax withholding, scheduled dispositions, and routine options exercises. Records still appear in the dataset but downstream automation can branch on the boolean to filter without dropping data.
• Quality-of-Signal scoring — Top-level signalQuality block with composite score 0-100 + tier (institutional-grade / high / moderate / background-noise) + rarity enum + confidence. The headline filter for high-conviction analysts and quants.
• 6-tier institutional severity taxonomy — Every record carries severity enum: informational / notable / elevated / significant / exceptional / critical. Standard institutional vocabulary across transactions, clusters, and issuer summaries.
• Multi-insider cluster coordination signals — Detected clusters carry coordinationSignals with coordinationStrength (0-1), executiveConcentration (high/medium/low/none), sameWeekParticipation count, hierarchySpread[] (CEO/CFO/Director hierarchy crossings), and timingCohesion (tight/moderate/loose) — cluster intensity beyond just member count.
• Per-issuer regime classifier — regimeState.current enum: accumulation-regime / liquidation-regime / governance-deterioration-regime / executive-confidence-recovery / insider-silence-breakdown / accelerating-distribution-regime / mixed-regime / no-regime. Hedge-fund-grade operational state vocabulary.
• Signal cards (dashboard-ready) — Each transaction emits up to 3 signalCards[] with type / severity / headline / confidence. Card types: rare-ceo-buy, senior-officer-buy, critical-officer-disposal, large-disposal, strategic-accumulator, founder-confidence-buy, liquidity-stress, extreme-deviation. Drop straight into Slack, newsletters, exec emails.
• Ownership delta intelligence — ownershipDelta block measures position-impact (holdingChangePct, priorHoldings, newHoldings, ownershipConviction) — distinct from raw dollar value. A 40% increase in holdings matters more than $1M.
• Context narrative — Paste-ready 3-line narrative per transaction (historicalSignificance / behavioralDeviation / timingInterpretation + summary). Generated deterministically from existing fields — no LLMs, no hallucinations.
• Direct SEC EDGAR access — Queries the official EDGAR full-text search index at efts.sec.gov for Form 4 filings with zero third-party dependencies
• Complete Form 4 XML parsing — Extracts both non-derivative transactions (stock purchases, sales, gifts) and derivative transactions (options exercises, conversions, expirations)
• Per-transaction decision layer — Every record carries direction, materiality (0-100 score + critical/high/medium/low band), unusualness, insiderRole (CEO/CFO/COO/Officer/Director/10% Owner with seniorityRank), decision.posture, recommendedAction, reviewSla (P0–P4), and escalationRecommendation
• Cluster detection — Surfaces ≥2 distinct insiders moving the same direction within a 30-day window, classified as officer-cluster-buy, officer-cluster-sale, broad-cluster-buy, broad-cluster-sale, or generic cluster-buy/cluster-sale
• Per-issuer state synthesis — issuerState block: sentiment (bullish/bearish/mixed/neutral), insiderPressureIndex (0-100 composite), transactionVelocity, governanceFootprint, watchStatus (no-action / monitor / attention-required / urgent / critical)
• Watchlist mode — Persists per-issuer state across runs. Emits temporalSignals, delta (new/changed/unchanged/recovered/degraded), stateTransition, historicalProfile (cumulative track record), regimeShift detection, and narrativeDelta (paste-ready exec sentences)
• Mode presets + persona-shaped routing — 7 modes (monitoring / cluster-detection / compliance / sentiment / officer-tracking / raw / auto), 7 personas (quant / fundamental-analyst / compliance / hedge-fund-research / retail-monitoring / corporate-ir / generic), 7 routing profiles bundling persona + trigger packs + SLA tightening
• Trigger packs — Pre-built deterministic rule bundles: cluster-buy, cluster-sale, executive-disposal, large-purchase, amendment-burst, distress-watch. Add custom rules via the triggers input
• Multi-issuer portfolio mode — Pass queries: ["AAPL", "TSLA", "NVDA"] to operate as a portfolio decision engine. Emits portfolioState, priorityQueue, immediateReviewQueue, and cross-issuer watchlistAnalytics
• Suite navigation — Every record carries actorGraph, nextActions[], improvementSuggestions[], and inlineEnrichmentHints pointing at sibling actors (sec-edgar-filing-analyzer, company-deep-research, finnhub-stock-data, congress-stock-tracker)
• Decision snapshot for compliance replay — Every record carries decisionSnapshot with inputsHash, rulesApplied[], and profileSnapshot so auditors can reproduce any decision deterministically
• KV mirrors — SUMMARY (full batch summary), OUTPUT (deterministic full output), MANIFEST (run manifest with internal-table version pins), ALERTS (immediate-review queue), COST_MODEL (per-outcome predicted spend)
• Output profiles — minimal / standard / full / llm shrink the field set without losing the decision primitives
• Flexible search — Search by company name ("Apple Inc."), ticker symbol ("AAPL"), CIK number ("320193"), or an insider's name
• Transaction type filtering — Narrow results to only purchases, sales, grants/awards, or options exercises
• Custom date ranges — Specify start and end dates in YYYY-MM-DD format, or rely on the default 30-day lookback window
• Minimum value threshold — Exclude small or insignificant transactions by setting a dollar floor
• SEC rate-limit compliance — Built-in throttling at approximately 8 requests per second with exponential backoff retry on 429 responses
• No API key required — All data comes from publicly available SEC EDGAR endpoints

---

Choose your mode

Modes are JOB-named — pick what you're doing, not what features to enable. The original mechanical names (monitoring / cluster-detection / etc.) remain as legacy aliases.

Mode	Job	Bundles
alpha-hunter	Find rare officer buys, strategic accumulation, high-quality bullish insiders	clusters + issuer state + delta + cluster-buy + large-purchase + distress-watch packs
governance-risk	Detect executive unloading, clustered exits, deterioration patterns	full output profile + delta + cluster-sale + executive-disposal + amendment-burst + distress-watch packs
compliance-surveillance	Full activity audit trail with replay-ready decision snapshots	full output profile + delta + every cluster + executive + amendment pack
executive-sentiment	C-suite directional behaviour only (CEO / CFO / COO / 10% Owner)	clusters + executive-disposal + large-purchase packs
distress-radar	Liquidity-pattern sells, accelerated disposition, emergency-behaviour signals	clusters + delta + cluster-sale + executive-disposal + distress-watch packs
monitoring (legacy)	Daily / weekly portfolio scans	clusters + issuer summary + batch insights
cluster-detection (legacy)	Cluster-only signal feeds	clusters only, no issuer summary
compliance (legacy alias)	Compliance review pipelines	full surface + delta + audit replay
sentiment (legacy alias)	Sentiment dashboards	clusters + issuer state + delta
officer-tracking (legacy alias)	CEO/CFO/COO/10% only	clusters + executive-disposal + large-purchase packs
raw	Power users — direct data access, no decision layer	no clusters, no summaries, raw transactions
auto	Auto-detect based on input shape	currently resolves to monitoring; full input-shape detection deferred

Choose your persona

Persona	Suppresses	Escalates on	Audience
quant	tax-withholding, small-grant	large-purchase, cluster-buy, cluster-sale	quant-research, algo-signal
fundamental-analyst	tax-withholding, small-grant, small-exercise	executive-disposal, cluster-buy, large-purchase	equity-research, analyst
compliance	(nothing — full surface)	executive-disposal, cluster-buy, cluster-sale, amendment-burst	compliance, legal, audit
hedge-fund-research	tax-withholding	cluster-buy, cluster-sale, distress-watch	hf-pm, hf-analyst
retail-monitoring	tax-withholding, small-grant, small-exercise, small-sale	cluster-buy, large-purchase	retail, newsletter
corporate-ir	small-grant	cluster-sale, executive-disposal	ir-team, cfo-office
generic (default)	tax-withholding	cluster-buy, cluster-sale, large-purchase, executive-disposal	general

Every persona has a documented ignoreThreshold. When alertQuality.urgency falls below that threshold for a record, escalationRecommendation.ignoreRecommendation is set to true so downstream automation can filter the noise without losing the data.

Choose your routing profile

Routing profiles bundle persona + trigger packs + SLA tightening + alert payloads in one input:

Profile	Persona	Trigger packs	SLA tightening
quant-monitor	quant	cluster-buy, cluster-sale, large-purchase	0h
compliance-officer	compliance	cluster-buy, cluster-sale, executive-disposal, amendment-burst	24h
hedge-fund	hedge-fund-research	cluster-buy, cluster-sale, distress-watch	12h
pe-operating	fundamental-analyst	executive-disposal, cluster-sale	0h
corporate-ir	corporate-ir	cluster-sale, executive-disposal, distress-watch	24h
retail-investor	retail-monitoring	cluster-buy, large-purchase	0h
generic (default)	generic	cluster-buy, cluster-sale, large-purchase	0h

Cohort-scale audits (50+ tickers, year-scale dates)

Apify's default 1h run timeout can be exceeded on large cohorts. Two patterns work:

1. Override the timeout per run — set timeoutSecs: 0 (unlimited) or timeoutSecs: 14400 (4h) in your run options. The actor's defaultRunOptions.timeoutSecs is set to 7200 (2h) so most cohort runs complete naturally.
2. Chunk by sector via Apify Schedules — split S&P 500 into ~11 GICS sector batches × ~50 tickers each. Each chunk completes in well under the timeout; merge downstream. Apify Schedules support multiple Tasks per schedule.

The runtime budget (runtimeBudgetSeconds, default 3000s) is a separate soft cap that emits partial results instead of letting Apify's hard timeout produce a silent failure. For very large cohorts, raise it.

---

When NOT to use this actor

Anti-use case	Use this sibling actor instead
You want all SEC filings (10-K, 10-Q, 8-K, DEF 14A) — not just Form 4	sec-edgar-filing-analyzer
You want full-text search across the EDGAR filing index with flexible filters	edgar-filing-search
You want congressional trading (Members of Congress, not corporate insiders)	congress-stock-tracker
You want stock price + fundamentals to combine with insider signals	finnhub-stock-data
You want a broader company intelligence brief that includes insider activity as one signal	company-deep-research
You want 13F holdings (institutional positions, not individual transactions)	sec-13f-holdings

Comparison vs alternatives

Capability	This actor (v2)	Bloomberg Terminal	OpenInsider	Form4 RSS feeds
Direct SEC EDGAR source	Yes	Indirect	Yes	Yes
Per-transaction decision layer (materiality, unusualness, escalation)	Yes	No	No	No
Cluster detection	Yes (deterministic)	Manual	Partial	No
Per-issuer state engine + watch status	Yes	Manual	No	No
Watchlist + cross-run drift + regime shift	Yes	Yes	No	No
Persona-shaped escalation (quant / compliance / IR / etc.)	Yes	No	No	No
Trigger packs + custom rules	Yes	Yes	No	No
Compliance audit replay (decisionSnapshot)	Yes	Limited	No	No
Cost	$0.002/transaction	$24,000+/year/seat	Free (web)	Free (web)

Insider trading data is informational. None of the outputs of this actor are investment advice.

How to use

Using the Apify Console

1. Go to the SEC Insider Trading Tracker page on the Apify Store.
2. Click Try for free to open the actor in your Apify Console.
3. Enter a company name, ticker symbol, or CIK number in the Search Query field (e.g., AAPL).
4. Optionally select a Transaction Type to filter by -- Purchase, Sale, Grant/Award, or Exercise of Options.
5. Optionally set Date From and Date To fields in YYYY-MM-DD format.
6. Optionally set a Minimum Transaction Value in dollars to exclude small trades.
7. Adjust Max Results if needed (default is 25, maximum is 200).
8. Click Start to run the actor.
9. When the run completes, view results in the Dataset tab and export as JSON, CSV, Excel, or XML.

Using the Apify API

cURL:

curl "https://api.apify.com/v2/acts/ryanclinton~sec-insider-trading/run-sync-get-dataset-items?token=YOUR_API_TOKEN" \
  -X POST \
  -d '{
    "query": "TSLA",
    "transactionType": "purchase",
    "dateFrom": "2025-06-01",
    "dateTo": "2025-12-31",
    "minTransactionValue": 100000,
    "maxResults": 50
  }' \
  -H "Content-Type: application/json"

JavaScript:

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_API_TOKEN' });

const run = await client.actor('ryanclinton/sec-insider-trading').call({
    query: 'TSLA',
    transactionType: 'purchase',
    dateFrom: '2025-06-01',
    dateTo: '2025-12-31',
    minTransactionValue: 100000,
    maxResults: 50,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items);

Python:

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")

run = client.actor("ryanclinton/sec-insider-trading").call(run_input={
    "query": "TSLA",
    "transactionType": "purchase",
    "dateFrom": "2025-06-01",
    "dateTo": "2025-12-31",
    "minTransactionValue": 100000,
    "maxResults": 50,
})

items = client.dataset(run["defaultDatasetId"]).list_items().items
for item in items:
    print(f"{item['insiderName']} - {item['transactionType']} - ${item['totalValue']:,.2f}")

---

Input parameters

Parameter	Type	Required	Default	Description
query	String	Yes	AAPL	Single-issuer query — company name, ticker, or CIK
queries	Array<String>	No	—	Multi-issuer portfolio — when set, takes precedence over query
transactionType	String	No	all	all / purchase / sale / grant / exercise
dateFrom	String	No	30 days ago	Start date (YYYY-MM-DD)
dateTo	String	No	Today	End date (YYYY-MM-DD)
minTransactionValue	Number	No	—	Dollar floor for transaction value
maxResults	Integer	No	25	Cap on transaction records (1–200). Cluster + issuer + batch records do NOT count against this cap.
mode	String	No	monitoring	See "Choose your mode" table
persona	String	No	generic	See "Choose your persona" table
routingProfile	String	No	generic	See "Choose your routing profile" table
systemMode	Boolean	No	false	One-flag stateful service — auto-enables clusters + issuer summary + batch insights + (with watchlistName) deltaMode
watchlistName	String	No	—	Names a per-watchlist named KV store for cross-run state
deltaMode	Boolean	No	false	When true (and watchlistName set), every transaction carries a delta block
includeClusters	Boolean	No	mode-driven	Detect ≥2 distinct insiders moving same direction within 30d
includeIssuerSummary	Boolean	No	mode-driven	Emit one issuer-summary record per ticker
includeBatchInsights	Boolean	No	mode-driven	Emit batch-summary record at run end
triggerPacks	Array<String>	No	profile-driven	Pre-built rule bundles (cluster-buy / cluster-sale / executive-disposal / large-purchase / amendment-burst / distress-watch)
triggers	Array	No	—	Custom rules: { name, when: { direction?, minTotalValue?, minClusterSize?, insiderRoleIn?, materialityLevelIn?, unusualnessAtLeast? }, severity? }
outputProfile	String	No	standard	minimal / standard / full / llm
includeAlertPayloads	Boolean	No	profile-driven	Write immediate-review queue + alerts to the run's ALERTS KV key

Example input:

{
    "query": "AAPL",
    "transactionType": "purchase",
    "dateFrom": "2025-01-01",
    "dateTo": "2025-12-31",
    "minTransactionValue": 100000,
    "maxResults": 50
}

Tips for best input:

• Use ticker symbols (e.g., TSLA, AAPL, MSFT) for the most precise results
• Use a CIK number if your company name search returns filings for the wrong entity
• Widen the date range for smaller companies that file infrequently -- try 90 or 180 days
• Combine transactionType: "purchase" with a high minTransactionValue to isolate significant insider buying signals
• Leave transactionType as all for a complete picture of insider activity

---

Output

The dataset emits multiple record types — branch on recordType to route them downstream:

recordType	What it represents	Per-run count
transaction	One Form 4 transaction with full decision enrichment	up to maxResults
cluster	A multi-insider cluster within a 30-day window	0 – N
issuer-summary	Per-issuer state synthesis	1 per distinct issuer
batch-summary	Run-level portfolio + priority queue + manifest	1 (or 0 in raw mode)
no-results	Empty-result explanation (no charge)	0 or 1
error	Run-failure record (no charge)	0 or 1

Example transaction record (v2):

{
    "schemaVersion": "2.0.0",
    "recordType": "transaction",
    "eventId": "a3f0e89b1c4d2e7f5a6b8c91",
    "filingDate": "2025-08-15",
    "companyName": "Apple Inc.",
    "ticker": "AAPL",
    "cik": "320193",
    "insiderName": "WILLIAMS JEFFREY E",
    "insiderTitle": "COO",
    "relationship": "Officer",
    "transactionType": "Purchase",
    "transactionDate": "2025-08-13",
    "sharesTraded": 15000,
    "pricePerShare": 178.52,
    "totalValue": 2677800,
    "sharesOwnedAfter": 489322,
    "directOrIndirect": "Direct",
    "direction": "buy",
    "insiderRole": { "isCeo": false, "isCfo": false, "isCoo": true, "isOfficer": true, "isDirector": false, "isTenPercent": false, "seniorityRank": 85, "label": "COO" },
    "materiality": { "score": 78, "level": "critical", "drivers": ["value:1M-10M", "role:senior-officer", "position-impact:3-10%", "ownership:direct"], "sharePctOfPriorHoldings": 3.2 },
    "unusualness": { "score": 65, "level": "high", "factors": ["officer-open-market-buy-uncommon", "very-large-dollar-amount"] },
    "decision": { "posture": "investigate", "confidence": 0.81, "confidenceLevel": "high", "drivers": ["senior-officer-buy", "high-materiality"], "whyThisMatters": "COO open-market buy of critical materiality — senior-officer purchases are uncommon and often correlate with management confidence.", "whyNow": "Pattern is unusual — officer-open-market-buy-uncommon; very-large-dollar-amount." },
    "decisionTrace": ["mode:monitoring", "persona:generic", "routing-profile:generic", "direction:buy", "role:coo", "role:officer", "materiality:critical", "unusualness:high", "ownership:direct"],
    "alertQuality": { "novelty": 0.65, "urgency": 0.97, "noiseRisk": 0.19 },
    "recommendedAction": { "actionId": "investigate-today", "label": "Investigate COO buy — critical materiality", "team": "analyst", "urgency": "today", "eta": "24h", "reason": "investigate (critical materiality, COO, persona=generic, profile=generic)", "contributingSignals": ["senior-officer-buy", "high-materiality", "materiality:critical", "role:coo"] },
    "reviewSla": { "recommendedReviewWithinHours": 4, "slaTier": "P1", "reason": "posture=investigate · role=COO · materiality=critical · profile=generic (tighten=0h)" },
    "escalationRecommendation": { "level": "immediate-review", "sla": "24h", "reason": "Escalated as immediate-review — investigate, critical materiality, persona=generic.", "audience": ["general"], "ignoreRecommendation": false },
    "summary": "AAPL — COO WILLIAMS JEFFREY E bought 15000 shares ($2678k) on 2025-08-13. Materiality critical (78/100). Posture: investigate."
}

Example cluster record:

{
    "schemaVersion": "2.0.0",
    "recordType": "cluster",
    "eventId": "c47b29d8e1a3f5b62c4d8e9f",
    "ticker": "TSLA",
    "cik": "1318605",
    "companyName": "Tesla, Inc.",
    "direction": "sell",
    "pattern": "officer-cluster-sale",
    "insiderCount": 4,
    "transactionCount": 9,
    "totalValue": 47200000,
    "spanDays": 12,
    "confidence": "high",
    "decisionPosture": "flag-disposal",
    "summary": "TSLA — officer-cluster-sale (4 insiders, 9 txns, $47200k) over 12d. Confidence high."
}

Example issuer-summary record:

{
    "schemaVersion": "2.0.0",
    "recordType": "issuer-summary",
    "ticker": "TSLA",
    "transactionCount": 12,
    "buyCount": 0,
    "sellCount": 9,
    "netInsiderFlow": -47200000,
    "issuerState": { "sentiment": "bearish", "insiderPressureIndex": 78, "insiderPressureBand": "high", "transactionVelocity": "elevated", "governanceFootprint": "high", "watchStatus": "urgent", "headline": "TSLA: urgent — sentiment bearish, pressure 78/100 (high), 12 txns / 1 clusters." },
    "triggersFired": [{ "name": "cluster-sale:officer-cluster", "severity": "high", "matchedSignals": ["cluster:officer-cluster-sale:4"] }],
    "decisionTrace": ["sentiment:bearish", "pressure-band:high", "watch-status:urgent", "velocity:elevated", "governance:high", "trigger:cluster-sale:officer-cluster"],
    "recommendedAction": { "team": "analyst", "urgency": "today", "label": "Investigate COO disposal — high materiality" },
    "reviewSla": { "slaTier": "P1", "recommendedReviewWithinHours": 4 },
    "escalationRecommendation": { "level": "immediate-review", "ignoreRecommendation": false }
}

Stable enum tokens

These enums are additive across minor versions — values can be added, never renamed or repurposed within a major version:

• recordType: transaction / cluster / issuer-summary / batch-summary / no-results / error
• direction: buy / sell / grant / exercise / gift / other
• materiality.level: critical / high / medium / low
• unusualness.level: high / medium / low
• decision.posture: investigate / flag-cluster / flag-disposal / monitor-sentiment / monitor / no-action
• escalationRecommendation.level: immediate-review / weekly-review / monthly-review / archive / ignore
• recommendedAction.team: compliance / ir / legal / finance / executive-review / analyst / quant / archive
• recommendedAction.urgency: immediate / today / this-week / this-month / no-action
• reviewSla.slaTier: P0 (1h) / P1 (4h) / P2 (48h) / P3 (1 week) / P4 (no SLA)
• issuerState.watchStatus: no-action / monitor / attention-required / urgent / critical
• issuerState.sentiment: bullish / bearish / mixed / neutral / unknown
• issuerState.insiderPressureBand: critical / high / elevated / normal / quiet
• cluster.pattern: officer-cluster-buy / officer-cluster-sale / broad-cluster-buy / broad-cluster-sale / cluster-buy / cluster-sale
• delta.type: new / changed / unchanged / recovered / degraded
• failureCategory: no-results / edgar-search-failure / xml-parse-failure / validation / rate-limit / timeout / transient-network / unknown
• severity (6-tier institutional taxonomy): informational / notable / elevated / significant / exceptional / critical
• behaviorProfile.profileType (14 archetypes): strategic-accumulator / conviction-buyer / opportunistic-buyer / crisis-accumulator / founder-confidence-buy / compensation-cycler / routine-options-exerciser / scheduled-seller / liquidity-manager / strategic-distributor / accelerated-disposal / governance-exit-pattern / pre-event-deleveraging / unclassified
• behaviorProfile.historicalBias: buy-heavy / sell-heavy / mixed / compensation-only / insufficient-data
• behaviorProfile.signalReliability: high / medium / low / cold-start
• intentClassification.primaryIntent (12 intents): discretionary-conviction-buy / discretionary-open-market-buy / opportunistic-accumulation / strategic-disposal / liquidity-stress-sale / coordinated-reduction / compensation-vesting / tax-obligation / scheduled-10b5-1 / estate-transfer / trust-reallocation / unclassified
• noiseClassification.noiseType: routine-compensation-cycle / scheduled-disposition / tax-withholding / options-exercise / low-materiality-grant / none / unclassified
• signalQuality.tier: institutional-grade / high / moderate / background-noise
• signalQuality.rarity: extreme / rare / uncommon / common / routine
• regimeState.current (8 regimes): accumulation-regime / liquidation-regime / governance-deterioration-regime / executive-confidence-recovery / insider-silence-breakdown / accelerating-distribution-regime / mixed-regime / no-regime
• coordinationSignals.executiveConcentration: high / medium / low / none
• coordinationSignals.timingCohesion: tight / moderate / loose
• ownershipDelta.direction: increased / decreased / unchanged / unknown
• ownershipDelta.ownershipConviction: high / moderate / low / unknown
• signalReputation.reputationTier: elite / high / moderate / low / cold-start
• signalVerdict.classification (9 values): high-conviction-bullish / bullish / speculative-bullish / governance-risk / distribution-risk / bearish / compensation-noise / informational / mixed-signal
• signalVerdict.actionability: immediate-review / weekly-review / monitor / archive / ignore
• signalVerdict.institutionalRelevance: high / medium / low
• signalLifecycle.freshness: fresh / recent / aging / stale
• signalLifecycle.signalDecayRisk: low / medium / high
• signalMaturity.stage: cold-start / emerging / developing / mature
• decisionReadiness.state: thesis-ready / partial-thesis / monitor / insufficient-data / noise
• expectationDeviation.deviationSeverity: extreme / high / moderate / low / none / first-sight
• behavioralDrift.directionalShift: bullish-to-defensive / defensive-to-bullish / accumulation-to-distribution / distribution-to-accumulation / stable / mixed / first-sight
• behavioralDrift.driftVelocity: accelerating / steady / decelerating / none / unknown
• governanceDynamics.executiveAlignment: aligned / fractured / mixed / unknown
• governanceDynamics.boardConfidence: rising / stable / falling / unknown
• governanceDynamics.boardVsExecutiveDirection: same / opposite / unknown
• modelIntegrity.classificationIntegrity: high / medium / low
• reviewEconomics.expectedSignalValue: high / medium / low / noise
• reviewEconomics.triageEfficiency: excellent / good / fair / poor
• signalObfuscationRisk.distributionFragmentation: high / medium / low / none
• signalObfuscationRisk.concealmentLikelihood: high / moderate / low / none
• portfolioExposureImpact.portfolioRiskShift: increasing / decreasing / stable / first-sight
• portfolioExposureImpact.concentrationConcern: high / medium / low / none
• confidenceCalibration.calibrationQuality: well-calibrated / overconfident / underconfident / cold-start / unknown
• confidenceCalibration.overconfidenceRisk: high / medium / low / unknown
• signalAttribution.observedOutcome: bullish-confirmed / bearish-confirmed / neutral / pending / unknown
• signalAttribution.signalEffectiveness: high / medium / low / pending / unknown
• leadershipConsensusModel.consensusState: aligned / softening / breaking-down / fractured / unknown
• leadershipConsensusModel.dominantFaction: executive / board / tenpct-owner / mixed / none
• manipulationResistance.gamingResistance: high / medium / low
• manipulationResistance.washPatternRisk: high / medium / low / none
• macroSensitivity.marketRegime: risk-on / risk-off / neutral / unknown
• humanFeedbackLoop.analystDisposition: validated / overridden / pending / unknown

Output field descriptions:

Field	Type	Description
filingDate	String	Date the Form 4 was filed with the SEC (YYYY-MM-DD)
companyName	String	Full legal name of the issuing company
ticker	String	Stock ticker symbol
cik	String	SEC Central Index Key for the company
insiderName	String	Name of the reporting insider (Last, First format)
insiderTitle	String	Official title -- CEO, CFO, VP, General Counsel, etc.
relationship	String	Insider role: Director, Officer, 10% Owner, Other, or a combination
transactionType	String	Transaction label -- Purchase, Sale, Grant/Award, Exercise/Conversion, Gift, Tax Withholding, etc.
transactionDate	String	Date the transaction was executed (YYYY-MM-DD)
sharesTraded	Number or null	Number of shares involved in the transaction
pricePerShare	Number or null	Price per share at time of transaction in USD
totalValue	Number or null	Total dollar value (sharesTraded x pricePerShare)
sharesOwnedAfter	Number or null	Total shares held by the insider after the transaction
directOrIndirect	String	Direct for personally held shares, Indirect for shares held through trusts or entities
filingUrl	String	Direct URL to the original Form 4 XML filing on SEC.gov
secUrl	String	URL to browse all Form 4 filings for this company on EDGAR

---

Use cases

• Investment research -- Screen for insider buying clusters as a bullish signal before earnings announcements or major catalysts
• Compliance monitoring -- Audit insider trading activity across portfolio companies to flag potential conflicts or suspicious timing
• Financial journalism -- Identify unusually large insider sales or purchases for investigative reporting
• Quantitative signals -- Build algorithmic trading models that incorporate insider transaction patterns as alpha factors
• Corporate governance analysis -- Track whether executives are increasing or decreasing their ownership stakes over time
• Venture and PE due diligence -- Check whether insiders at a target company are buying or selling ahead of a potential deal
• Academic research -- Gather structured insider trading datasets for financial economics studies and market efficiency research
• Alert and notification systems -- Feed results into Slack, email, or webhook pipelines to get notified when specific insiders trade
• Portfolio risk management -- Monitor insider selling activity across your holdings as an early warning indicator

---

API & Integration

The SEC Form 4 API is exposed via the standard Apify Run-Sync endpoint, returning structured JSON for any insider-trading monitoring or alert workflow.

Python

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")
run = client.actor("ryanclinton/sec-insider-trading").call(run_input={
    "query": "MSFT",
    "transactionType": "sale",
    "maxResults": 100,
})
for item in client.dataset(run["defaultDatasetId"]).list_items().items:
    print(f"{item['insiderName']}: {item['transactionType']} {item['sharesTraded']} shares @ ${item['pricePerShare']}")

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_API_TOKEN' });
const run = await client.actor('ryanclinton/sec-insider-trading').call({
    query: 'MSFT',
    transactionType: 'sale',
    maxResults: 100,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
items.forEach((item) => console.log(`${item.insiderName}: $${item.totalValue?.toLocaleString()}`));

cURL

curl "https://api.apify.com/v2/acts/ryanclinton~sec-insider-trading/run-sync-get-dataset-items?token=YOUR_API_TOKEN" \
  -X POST \
  -d '{"query": "MSFT", "transactionType": "sale", "maxResults": 100}' \
  -H "Content-Type: application/json"

Platform integrations

• Zapier -- Trigger workflows in thousands of apps when insider trades are detected
• Make (Integromat) -- Build visual automation scenarios with insider trading data
• Google Sheets -- Export results directly to a spreadsheet for collaborative analysis
• Slack -- Send real-time notifications when high-value insider trades are filed
• Webhooks -- Trigger downstream actions when a run completes
• Scheduled runs -- Set up daily or weekly monitoring of insider trading activity

---

Use in Dify

Drop this actor into Dify workflows via the Apify plugin's Run Actor node. Each run returns scored, classified, and routed Form 4 activity as structured JSON — recordType (transaction / cluster / issuer-summary / batch-summary), escalationRecommendation.level (immediate-review / weekly-review / monthly-review / archive / ignore), decision.posture (investigate / flag-disposal / monitor-sentiment / monitor / no-action), and reviewSla.slaTier (P0 / P1 / P2 / P3 / P4) — plus the decisionCards[] and priorityQueue[] your downstream node branches on. Bloomberg or OpenInsider pointed at the same Form 4 filings return raw transactions; this returns decisions.

• Actor ID: ryanclinton/sec-insider-trading
• Sample input (compliance-officer routing across a portfolio):

{
    "queries": ["TSLA", "NVDA", "AAPL", "MSFT", "META"],
    "mode": "compliance",
    "routingProfile": "compliance-officer",
    "watchlistName": "tech-megacaps-weekly",
    "systemMode": true
}

Branching example — a Dify if/else node routing on escalationRecommendation.level:

IF escalationRecommendation.level == "immediate-review"
    → Send Slack alert to #compliance-urgent
    → Open Jira P0 ticket via HTTP node
ELIF escalationRecommendation.level == "weekly-review"
    → Append to weekly review queue (Notion / Airtable)
ELIF cluster.category == "risk" AND cluster.confidence == "high"
    → Run sec-edgar-filing-analyzer for the same ticker
ELSE
    → Archive

Opt-in modes Dify workflows can leverage:

• mode: "monitoring" (default) — every run emits clusters + issuer summaries + batch insights. Pair with a daily Dify schedule.
• mode: "compliance" — full decision surface + delta tracking. Pair with watchlistName + systemMode: true for cross-run drift detection.
• mode: "officer-tracking" — narrows to CEO/CFO/COO/10% owner activity. Useful when the Dify workflow only routes on senior-officer signals.
• outputProfile: "llm" — strips diagnostic blocks, keeps decisionCards, escalationRecommendation, decision, materiality, summary, headline, oneLine. The LLM-friendly subset for agent consumers that branch on enums and quote prose verbatim.

The decisionCards[] array, priorityQueue[] array, nextActions[] array, and improvementSuggestions[] array are usable verbatim — every string field (title, shortReason, summary, narrativeDelta.summary, recommendedAction.label) is paste-ready into the next node's prompt or into a Slack/email body. No LLM rewriting required.

For agent consumers, branch directly on recommendedAction.executionHint.type — six stable values (investigate-officer-buy / review-officer-disposal / monitor-cluster / monitor-sentiment / archive / no-action) plus executionHint.targetTicker + executionHint.targetCik give a runtime-routable surface without parsing any prose.

---

How it works

The actor follows a multi-step pipeline to search, retrieve, parse, filter, and output insider trading data from SEC EDGAR.

1. Search EDGAR -- Sends a paginated query to the SEC EDGAR full-text search index (efts.sec.gov/LATEST/search-index) filtering for Form 4 filings within the specified date range. Retrieves up to 3x the requested maxResults to allow for post-filtering.
2. Build XML URLs -- Extracts the accession number and filename from each search hit ID, then constructs the direct URL to the Form 4 XML filing on SEC.gov Archives.
3. Fetch Form 4 XML -- Downloads each raw Form 4 XML filing, respecting the SEC's rate limits with 120ms pauses between requests and exponential backoff on 429 responses.
4. Parse XML -- Uses fast-xml-parser to parse the XML into a structured JavaScript object, handling both single and multiple reporting owners, non-derivative transactions, and derivative transactions.
5. Extract transactions -- Pulls out every reportable transaction including shares traded, price per share, total value, post-transaction holdings, insider identity, relationship, and ownership type.
6. Apply filters -- Filters transactions by the requested transaction type (purchase, sale, grant, exercise) and minimum transaction value threshold.
7. Sort and limit -- Sorts results by filing date descending, then by total value descending, and trims to the requested maxResults count.
8. Output -- Pushes the final structured results to the Apify dataset for download or API retrieval.

Query (ticker/name/CIK)
           |
           v
  +-----------------+
  | EDGAR Full-Text |
  | Search Index    |
  +-----------------+
           |
     Form 4 hits
           |
           v
  +-----------------+
  | Build XML URLs  |
  | from accession  |
  +-----------------+
           |
           v
  +-----------------+
  | Fetch Form 4    |
  | XML (rate-      |
  | limited)        |
  +-----------------+
           |
           v
  +-----------------+
  | Parse XML &     |
  | extract trades  |
  +-----------------+
           |
           v
  +-----------------+
  | Filter, sort,   |
  | limit results   |
  +-----------------+
           |
           v
     Apify Dataset
     (JSON/CSV/Excel)

---

Performance & cost

Metric	Value
Memory	256 MB
Typical run time (25 results)	15--45 seconds
Typical run time (100 results)	45--120 seconds
Typical run time (200 results)	2--4 minutes
Compute units (25 results)	~0.005 CU
Compute units (100 results)	~0.02 CU
Compute units (200 results)	~0.04 CU
Estimated cost (25 results)	~$0.005
Estimated cost (200 results)	~$0.04
Free tier capacity	~1,000 runs/month on Apify's $5 free credit

Run time depends primarily on the number of Form 4 filings fetched from SEC EDGAR and the SEC's rate-limiting behavior. The actor does not use a browser, so memory and CPU usage remain minimal.

---

Limitations

• SEC rate limits -- The SEC imposes a 10 requests-per-second limit. The actor throttles to approximately 8/sec, but heavy EDGAR traffic or concurrent users may trigger 429 responses that slow processing.
• Form 4 coverage only -- This actor searches Form 4 filings specifically. It does not cover Form 3 (initial ownership statements), Form 5 (annual reports of changes), or 13D/13G (beneficial ownership filings).
• XML parsing edge cases -- A small number of Form 4 filings use non-standard XML structures, footnote-only values, or amended formats that may result in null fields for shares, price, or value.
• Search index latency -- While most filings appear in the EDGAR search index within minutes, some may take several hours to be indexed after submission.
• Name matching -- EDGAR full-text search does fuzzy matching, which can sometimes return results for similarly named companies. Use ticker symbols or CIK numbers for precise targeting.
• No historical backfill -- The EDGAR search index has practical limits on how far back paginated results extend. For deep historical analysis, consider using the EDGAR Filing Search actor.
• 200 result cap -- A single run returns a maximum of 200 transactions. For larger datasets, run multiple queries with narrower date ranges.

---

Responsible use

• Public data only -- This actor accesses publicly available SEC EDGAR filings. It does not scrape private, authenticated, or restricted data sources.
• SEC fair access compliance -- The actor respects SEC EDGAR's rate limits and identifies itself with a proper User-Agent header, following SEC guidelines for automated access.
• Not investment advice -- Insider trading data is informational. It should not be treated as a recommendation to buy, sell, or hold any security. Always do your own research and consult a financial advisor.
• Respect filing context -- Insider sales do not necessarily indicate negative sentiment. Many sales are pre-planned under Rule 10b5-1 plans, related to tax obligations, or part of portfolio diversification.
• Privacy considerations -- While Form 4 data is public, use insider names and transaction details responsibly and in accordance with applicable laws and your organization's policies.

---

FAQ

Q: Is this a SEC Form 4 API? A: Yes. SEC Insider Trading Tracker is a SEC Form 4 API — a developer-first insider trading API for monitoring, parsing, classifying, and routing Form 4 filings in real time. Output is structured JSON with deterministic enums (signalVerdict.classification, decisionReadiness.state, severity), suitable for Slack alerts, PagerDuty rules, Zapier filters, Dify workflows, and AI-agent tool calls. Pair with Apify Schedules for live SEC Form 4 monitoring.

Q: Does this actor require an API key or SEC registration? A: No. The actor queries the public SEC EDGAR full-text search index and parses publicly available Form 4 XML filings. No API key, registration, or authentication is needed.

Q: What is a Form 4 filing? A: Form 4 is the SEC filing that corporate insiders -- officers, directors, and shareholders who own 10% or more of a company's stock -- must submit within two business days of buying, selling, or otherwise transacting in their company's securities.

Q: What transaction types can I filter by? A: You can filter by Purchase (open-market buys), Sale (including tax withholding dispositions), Grant/Award (equity compensation grants), or Exercise (options exercises and conversions). Select "all" to see every transaction type including gifts, dispositions, and swaps.

Q: How current is the data? A: The actor queries the live SEC EDGAR search index, so filings appear as soon as the SEC processes them -- typically within minutes to hours of submission. By default, the actor searches the last 30 days.

Q: Why are some transactions missing price or value data? A: Some Form 4 filings report transactions at $0 (such as gifts, certain grants, or tax withholding at no cost) or omit price data entirely. The actor returns null for fields not present in the original XML filing.

Q: Can I search for a specific insider by name? A: Yes. Enter the insider's name in the query field. EDGAR's full-text search indexes insider names from Form 4 filings, so searching for "Elon Musk" or "Tim Cook" will return their filings.

Q: What is the difference between Direct and Indirect ownership? A: Direct ownership means the insider personally holds the shares. Indirect ownership means the shares are held through a trust, family member, partnership, LLC, or other entity where the insider has beneficial interest.

Q: How does the actor handle derivative transactions like options? A: The actor parses both the non-derivative table (common stock transactions) and the derivative table (options, warrants, convertible securities) from each Form 4 filing. Derivative transactions include the security title in the transaction type field (e.g., "Exercise/Conversion (Stock Option)").

Q: Can I schedule this actor to run automatically? A: Yes. Use the Apify platform's Schedules feature to run the actor daily, weekly, or at any custom interval. Combine with webhooks or Slack integration for automated alerts.

Q: What happens if the SEC rate limits the actor? A: The actor handles 429 (Too Many Requests) responses with exponential backoff -- waiting 2, 4, then 8 seconds between retries. It will automatically recover and continue processing without losing data.

Q: Why does the actor fetch more filings than my maxResults setting? A: The actor fetches up to 3x your requested maxResults from the EDGAR search index, then filters and deduplicates the parsed transactions. This over-fetching ensures enough results survive filtering by transaction type and minimum value.

Q: How is this different from the SEC EDGAR Filing Analyzer actor? A: The SEC EDGAR Filing Analyzer searches across all filing types (10-K, 10-Q, 8-K, etc.) and extracts filing metadata. This actor focuses exclusively on Form 4 insider trading filings and deeply parses the XML to extract individual transaction details including shares, prices, and values.

---

Related actors

Actor	Description
SEC EDGAR Filing Analyzer	Search and analyze all SEC EDGAR filing types -- 10-K, 10-Q, 8-K, and more
EDGAR Filing Search	Full-text search across the complete EDGAR filing index with flexible filters
Congressional Stock Trade Tracker	Track stock trades disclosed by members of the U.S. Congress
Finnhub Stock Market Data	Real-time and historical stock market data including quotes and company profiles
Company Deep Research Agent	Comprehensive company intelligence from multiple public data sources
OFAC Sanctions Search	Search the OFAC Specially Designated Nationals list for sanctioned entities