finbly
Explore
finbly
finbly
Multi-agent investment research

Know what changed.
Know if the thesis changed.

Finbly coordinates specialised agents, live-data sensors and deterministic analytical engines around public companies. They monitor filings, earnings calls, news, institutional activity, relationships and valuation signals—then turn the change into cited, explainable investment intelligence.

What changedWhy it mattersWhat to watch
909companies covered
~471kverified filing facts
2,792change triggers watched
270quarterly datasets
2026-07-03latest period covered
How Finbly works

One research system, not a collection of chatbots

Language models interpret evidence. Deterministic engines own the calculations. Stored outputs keep the record available after the conversation ends.

01Sensor

Observe the evidence

Filings, prices, earnings calls, news, Form 4s, 13Fs, estimates and short interest arrive on a published cadence.

02Agent

Interpret the change

Specialists analyse only bounded evidence: calls, guidance, risk factors, news and the evolving bull/bear case.

03Engine

Measure the impact

Deterministic valuation, scorecard, relationship and trigger engines calculate the parts that should not be guessed.

04Insight

Show what matters

The dossier preserves what changed, why it may matter and the evidence an investor should watch next.

Recent system work

Completed analyses, read from the outputs agents actually stored

This is not a sample feed. If a producer database is unavailable, its activity disappears rather than being replaced with a marketing placeholder.

Research that compounds

A durable record around every company

Finbly’s advantage is the accumulated structure, history and evidence—not dependence on one underlying model.

Specialist coverage

Count completed coverage, not imaginary agents

Each number comes from the database written by that specialist role. A configured role with no stored output counts as zero.

0
Filing extraction
Transcribe a filing into canonical, gated company financial data.
895
Earnings-call analysis
Brief each call and surface guidance, difficult Q&A, new risks and monitorable triggers.
0
News briefing
Classify news and explain the current company tape without inventing a catalyst.
265
Filing-risk classification
Extract and classify the company's filed risk factors.
208
Guidance extraction
Extract structured forward guidance from earnings materials.
232
Risk-pulse monitoring
Compare current news with filing-derived risks and surface changed pressure.
235
Thesis synthesis
Maintain grounded bull, bear and watch cases from the assembled company substrate.
217
Filing comparison
Identify added, removed and materially reworded disclosures between filings.
226
Management scorecard
Track explicit management commitments and resolve delivery over time.
299
Supply-chain intelligence
Interpret dated supplier and customer events through reviewed company relationships.
11
Competitive intelligence
Interpret dated rival developments within primary-source-reviewed competition arenas.
287
Insider activity interpretation
Distinguish open-market purchases, declared trading-plan sales and unclassified sales without inferring motive.
261
Valuation interpretation
Explain the expectations embedded in price and where deterministic valuation methods disagree.
300
Analyst estimate revisions
Track like-for-like changes in consensus EPS, revenue, price targets, rating distributions, upgrades and downgrades.
161
Short-positioning interpretation
Identify material changes in reported short positioning and distinguish positioning trend from a mechanical squeeze setup.
288
Regulatory event monitoring
Monitor explicit regulatory stages and connect them to the company's filing-derived regulatory exposure without deciding legal outcomes.
782
Fundamental credit capacity
Assess filing-derived debt burden, debt-service capacity and near-term refinancing pressure, with explicit credit-event context.
252
Evidence-linked bull/bear debate
Preserve the strongest opposing investment cases and identify the evidence that would resolve them.
233
Evidence-bound downside pre-mortem
Construct distinct company-specific paths to a roughly 30% drawdown and attach measurable early-warning indicators.
0 tokens
Verification engines
SEC XBRL comparison, arithmetic validation and price-sanity checks independently test the research substrate.
Why Finbly instead of a blank chat

It works before the investor knows what to ask

01

Persistent company memory

Filings, calls, relationships, trigger state and prior conclusions remain connected across reporting periods.

02

Change, not repetition

Scheduled sensors refresh the evidence and material changes update the living record instead of generating the same research from scratch.

03

Evidence over confidence

Facts, estimates and interpretations stay distinguishable, with source dates, freshness and evidence status exposed for inspection.

Primary-source foundationImportant financial facts are transcribed from filings, schema-gated, reconciled and independently checked against SEC XBRL anchors.
Bounded AI workModels classify, extract and synthesise targeted evidence. They do not own authoritative financial calculations or silently fill missing facts.
Visible evidenceThe Research Agents page exposes mandates, coverage and latest dated work without exposing internal model controls.
Building toward personal thesis monitoring

Today: a living company research system. Next: intelligence prioritised to the investor’s own holdings and thesis.

Saved portfolios, thesis memory, filing diffs and a management promise scorecard will turn the existing company substrate into a personal Thesis Guardian.