Graph AnalysisTemporalGeospatialLLM-native

Complete data integration.
Deep visualisation.

Mnemosyne Intelligence Suite unifies graph analysis, temporal investigation, geospatial mapping, and LLM-assisted reasoning in one integrated platform — built for team investigations, deployed on your own infrastructure.

Operation VeneeringCross-border trade fraud network4 high-risk6 countries
ForceRadial
ownscontrolsEUR 2.1M22 callsemploysEUR 550KsubsidiarytransfersUSD 340Kmanages!KA. MelmotteDombey & Son!Melmotte HoldingsCrypto WalletBank Acc #1Bank Acc #2J. ChuzzlewitNS. FledgebyVholesVeneering CapitalWegg TradingZurichCyprusLondonPodsnap LogisticsJarndyce Trust
person
org
account
location
Louvain: 4 communities detectedEUR 3.2M traced1 single point of failure
Team-based investigations
Graph + temporal + geospatial
LLM as analytical partner
Audit-logged by design
On-prem, hybrid-ready

Intelligence analysis needs more than link charts

A link chart shows who connects to whom. The questions that actually move an investigation — who controls the network, how it behaves over time, where its activity concentrates — need a typed model underneath the picture, not just the picture.

01

One actor, many records

The same person, account, or location arrives under a different key in every source — a phone number here, a name there, an IBAN somewhere else. Without a shared semantic model they stay separate rows, and the links that cross sources are never drawn.

Resolved in a shared ontology — one object per real-world entity, every source mapped onto it.

02

Influence is computed, not eyeballed

Who brokers between two groups, which node holds a cluster together, where the network splits if one actor is removed — these are structural facts of the whole graph. A link chart shows edges; it cannot rank centrality, detect communities, or expose the single points of failure.

Graph algorithms rank centrality, detect communities, and flag articulation points across the full network.

03

The signal is in time and place

When events happened, in what order, and where they originated often matters more than the connection itself. Treated as plain table columns, sequence and location can't be replayed, windowed, or correlated — so bursts, hand-offs, and co-location patterns stay buried.

Temporal playback and geospatial context make sequence and location first-class, queryable dimensions.

Data to intelligence in three steps

Import from any source. Analyse with algorithms and AI. Deliver structured intelligence products.

1

Import

Multi-source ETL

CSV, JSON, REST APIs, STIX 2.1 bundles. Schema auto-detection. Three-stage entity resolution across sources.

2

Analyse

Graph + Temporal + Geo

Centrality, communities, structural analysis. Temporal playback. Geospatial mapping. LLM reasoning. All views synchronized.

3

Deliver

Intelligence products

Briefs, network reports, entity profiles, risk assessments. Every finding traceable to source.

Integration Architecture

One semantic layer between your data and your decisions

Many sources converge into a single ontology — the narrow waist where objects, properties and links are defined. Analytics and applications fan back out from that one truth.

Visualization & Apps
Analytics & Logic
Ontology
Data Sources
Visualization & AppsConsumption

Dashboards, reports, alerts, workflows and agents — every product reads from the same ontology, so what one team sees, every team sees.

Analytical capabilities

Six capabilities operating on a shared ontology. Each handles a different analytical dimension. Results propagate across all views.

Graph Analysis

Networks

Centrality, communities, bridges — the structure behind the data.

Temporal Intelligence

Time

Playback, burst detection, change points — networks in motion.

Geospatial Mapping

Place

Live maps with heatmaps and trajectories, wired to the graph.

LLM Integration

AI partner

An LLM that reads the graph directly through structured tool calls.

Pattern Detection

Signals

Laundering structures, call-response chains, propagation patterns.

Data Ingest

Sources

Multi-source ingest with entity resolution into one semantic model.

Deep dive

Networks, decoded

Centrality metrics rank actors across multiple definitions — immediate reach, bridging importance, structural uniqueness. Modularity-based community detection reveals cluster structure at multiple scales; articulation points and bridges pinpoint the single links whose removal fractures the network.

  • Centrality across multiple definitions — from direct reach to structural influence
  • Modularity-based community detection, hierarchical
  • Articulation points, bridges, structural holes, latent-link prediction
MelmotteMerdleTulkinghornHoldingsDombeyFledgebyCryptoVeneeringBank #1ChuzzlewitWeggVholesWegg Trd.CyprusZurichLondonJarndyceBank #2PodsnapPodsnap Jr.Direct connectionsSecondary networkPeriphery

LLM assisted network analysis

Each mode builds on the previous. Chat discovers, Enrich refines, Anomaly surfaces, Report documents, Agent investigates.

Conversational Analysis

CHAT MODE

Ask questions in natural language. The LLM queries nodes, runs algorithms, computes paths, and explains results — all through structured MCP tool calls. Every answer is grounded in graph data.

Example: "Who bridges the financial and logistics clusters?"
Graph queriesAlgorithm resultsNatural language explanations
Analytical Factory

Design new analysis by describing it

The built-in algorithms are only the starting point. Describe an analytical idea in plain language and the platform assembles it into a graph algorithm that runs against the ontology. It composes the computation, exposes its parameters, and keeps the result as a building block you can reuse and refine.

  • State the metric or pattern you are after and the platform turns it into a concrete graph computation.
  • Every composed algorithm is inspectable and parameterised, never a black box.
  • Save what works as a reusable analytical building block, ready for the next investigation.
Analytical Factorygraph-native
ComposeRunSave

Find the intermediaries quietly moving value between the financial core and the shell-company network, and flag the ones whose role grew right before the latest transfers.

Analytical chain

1community detectionmodularity-based · core, shell, logistics
2betweenness centralityrestricted to core → shell paths
3burst + change-pointper actor transaction timeline
4edge betweenness + articulationisolate the actual conduits

Query results

1
A. Melmotte0.94

Highest cross-boundary betweenness; activity spiked just before the transfers.

2
Melmotte Holdings0.78

Articulation point linking the financial core to the shell layer.

Describe an analysis, or refine the composed algorithm...
Compose
Autonomous Actors

Investigators that live in the graph

1
The goal

Scale an analyst's attention across the whole graph.

One person cannot chase every lead in a large network by hand. The search has to run on its own and in parallel, and still stay controllable and traceable.

2
The concept

Investigators that live inside the graph.

Small autonomous actors sit on the graph itself. Each is bound to a single element (a person, a vessel, an event) and carries an assignment. They range from ready-made patterns for familiar cases to investigators that plan a multi-step approach on their own.

Bounded investigators pursue a clear goal and finish: find the hidden coordinator, trace the link from one entity to another.

Continuous monitors watch their territory and raise the alarm the moment something breaks the pattern.

3
How it runs

From a defined assignment to an autonomous strategy.

  1. 1

    Each actor receives a defined assignment.

  2. 2

    It works out its own strategy to interpret and analyse the data model.

  3. 3

    State awareness and communication between actors, so each builds on what the others surface.

4
The result

A trail in the graph, not an answer in a chat window.

Connections drawn, paths marked, findings left where they belong. Each one stays open to inspection and tied back to the steps that produced it.

What you deliver

Traceable, structured outputs for analysts, case officers, or regulatory authorities.

Intelligence Briefs

Structured reports from graph data. Executive summary, findings, network analysis, timeline, recommendations.

Network Reports

Community structure, centrality rankings, bridges, structural holes. Visual maps with algorithm overlays.

Entity Profiles

Per-entity dossiers: properties, relationships, temporal activity, geospatial footprint, risk assessment.

Risk Assessments

Entity and network-level scoring based on centrality, community position, anomaly flags, temporal patterns.

Live exploration view

The working surface behind every output

All four deliverables read from the same live, typed graph. Analysts explore entity aggregates, drill into communities, and validate findings against the raw stream — every export carries a verifiable lineage back to source records.

placed_by · 26received_by · 26funds_sent_by · 14funds_received_by · 14crossed_by · 13files_flight · 4crossing_vehicle · 13via_tower · 26originates_from · 14crossing_at · 13departs/arrives · 8aircraft_flies · 4communicates_with · 17sends_funds_to · 8PERSON · 12★ CENTRAL · person-centricCOMMUNICATION · 26HAWALA · 12BORDER · 13VEHICLE · 7FLIGHT · 4AIRCRAFT · 3PLACE · 9
Mnemosynelive graph
● stream live

Ready to see it in action?

Graph analysis, temporal playback, and LLM reasoning — in one browser-native environment.

Get in touch

Whether you are evaluating intelligence analysis tools, consolidating fragmented workflows, or exploring LLM-assisted investigation — describe your use case.

Tailored walkthroughFor your investigation domain
Deployment optionsBrowser, Electron, air-gapped
Early accessPriority onboarding available

This website uses cookies and similar browser storage, along with selected third-party services. The site remains usable without consent. Details in the privacy policy.

Privacy