AI early-warning system
for financial instability.
Phoring helps macro and risk teams detect early signs of instability. It turns geopolitical events, policy shifts, market context, and public signals into source-grounded risk scenarios.
From signals and evidence
to early-warning scenarios.
Signal Ingestion
Ingests public, policy, and market sources, extracts entities and relationships, and builds a traceable knowledge graph for instability monitoring.
Scenario Simulation
Deploys agent personas into synthetic environments to test how geopolitical and policy shocks could propagate across narratives and risk drivers.
Risk Scenarios
Generates source-cited risk scenarios with confidence-scored reports and early-warning alerts. Every claim traces back to evidence.
Four-stage pipeline.
Signals in, risk scenarios out.
Graph Build
Ingest public, policy, and market sources (PDF, Markdown, or text). Phoring extracts entities and relationships, then builds an evidence-grade knowledge graph in Zep.
Agent Setup
LLM-generated agent profiles reflect institutions, stakeholders, and narratives from the evidence layer, with scenario-specific behaviors and constraints.
Simulation
OASIS runs agents through synthetic market and public-discourse environments. Actions, reactions, and narrative shifts unfold across configurable rounds.
Intelligence Report
The Report Agent synthesizes graph evidence, web intelligence, and simulation outcomes into source-cited risk scenarios with confidence scoring and alerts.
Built for financial risk teams
monitoring instability and policy shocks.
Policy Shock Assessment
Assess how regulatory, fiscal, and central-bank actions could ripple across sectors, institutions, and narratives.
Market Narrative Shifts
Track how market narratives and risk sentiment shift around events and policy signals, with evidence-backed scenarios.
Financial Instability Monitoring
Monitor early signs of stress across markets, funding conditions, and sentiment to trigger early-warning scenarios.
Scenario-Based Risk Reporting
Produce committee-ready reports that link scenarios to sources, assumptions, and confidence scores.
Geopolitical Risk Tracking
Track geopolitical events, sanctions, and conflict signals to surface downstream financial exposure.
Contagion Pathway Analysis
Model how shocks propagate across sectors, jurisdictions, and narratives to identify spillover pathways.
Different by architecture,
not by hype.
Not a chatbot.
A scenario intelligence system.
Phoring doesn't answer from a single prompt. It runs multi-agent simulations to explore how policy and geopolitical shocks could ripple across markets and narratives.
Not generic summaries.
Grounded in evidence.
Every scenario traces back to documents, policy signals, market context, and web intelligence. Evidence quality directly shapes the output.
Not a black box.
Source-cited and scored.
Reports include inline references [1][2][3] to specific sources. Confidence scoring reflects evidence strength, not certainty.
Not a single perspective.
Multi-agent, multi-model.
Optional multi-model consensus validation runs independent models to check coherence, precedent, and risk framing before adding a consensus note.
Transparent by design.
Every step is auditable.
LLM-driven ontology extraction → entity and relationship mapping → Zep Cloud storage → downstream querying for agent profiles, simulation context, and report generation.
Entity-scoped queries via Serper and NewsAPI. Articles scraped and processed up to 4,000 characters each. Social content sourced via Google Search indexing.
OASIS framework spawns synthetic environments where LLM-generated agents interact based on assigned personas, stances, and behavioral parameters. Results stream in real time.
ReACT loop pulls from knowledge graph, web intelligence, and simulation data. Claims are backed by inline numbered references with a full sources section.
Each report section tagged [HIGH], [MEDIUM], or [LOW] based on independent tool-sourced data points. Reflects evidence density — not a guarantee of accuracy.
Optional. Up to 3 independent LLM validators score predictions on coherence, precedent, and risk. A consensus summary is appended to the final report.
Upload documents,
define your scenario.
Drop your source files and describe the scenario you want to simulate. Phoring handles the rest.
Drop files here or click to browse
PDF · MD · TXT — up to 50 MB
Start detecting
instability early.
Upload your sources, define a risk scenario, and let Phoring generate source-grounded alerts and confidence-scored reports.