LOADING
TACITUS is the governed data layer for conflict-aware AI. We transform scattered dispute traces into structured knowledge graphs that machines can query and reason over — with temporality, causality, and provenance preserved.
“A map and a postcard both show the same territory. Only one helps you find your way. TACITUS is the map.”
Large Language Models are fundamentally blind to the dynamics that drive disputes. They collapse time, confuse correlation with causation, and cannot trace facts to sources.
No Temporality
Embeddings flatten time into noise. "Promise broken" and "trust lost" become tokens, not a causal chain.
No Causality
LLMs learn correlations, not cause-effect. "Before" is not "caused." Fatal for conflict analysis.
No Provenance
Facts cannot be traced to sources. In legal, HR, and diplomatic contexts, this is a dealbreaker.
We don't try to make LLMs understand conflict. We create a structured knowledge layer that makes conflict data legible, queryable, and auditable.
DAG-Preserved Temporality — causal chains intact
Explicit Causal Links — CAUSED(A,B), not correlation
Full Provenance Chains — every fact traces to source
Deterministic Reasoning — auditable, reproducible, zero hallucination on keyed facts
Four stages transform scattered conflict traces into machine-queryable intelligence.
Emails, documents, chat logs, transcripts — any unstructured conflict trace.
Identify actors, claims, events, and evidence. LLMs used for extraction only, not judgment.
Agentic Conflict Ontology: type classification, relation binding, temporal ordering, provenance linking.
Graph-native storage. API-queryable. LLM-groundable. Audit-ready. Deterministic.
Purpose-built for conflict reasoning. No hallucinations. Full auditability.
Fills gaps in conflict history with statistically probable but factually incorrect data.
Treats all past interactions as a flat vector, losing the causal chain of events.
Impossible to audit why a specific negotiation tactic was chosen.
Anchors responses in verified structural memory nodes. 0% hallucination rate on keyed facts.
Understands causality. Knows that Event A caused Conflict B, regardless of context window size.
Trace every output back to the specific memory node and source document.
Built on the Conflict Graph
Products powered by the same ontological structure. Analysis, preparation, and decision support — grounded in verified facts.
All tools are in active development
Structure cases, map claims over time, and surface potential paths forward.
Early access · Iterating with design partners
Hybrid mediator for scenario exploration and structured briefings. AI reasoning grounded in your conflict graph.
In development — targeted launch: Q2 2026.
Stress-test strategies, narratives, and policy positions against structured conflict data before they hit the real world.
Early access · Iterating with design partners
Wherever parties disagree — disputes, grievances, negotiations, governance friction — TACITUS brings the same rigor to conflict data that financial systems bring to transactions. One architecture, infinite applications.
Grievance structuring, pattern detection, compliance
Evidence mapping, provenance chains, discovery
Obligation tracking, breach analysis, M&A
Stakeholder mapping, interest analysis, scenarios
Position tracking, narrative shifts, coalitions
Relationship mapping, interest surfacing, mediation
A simplified, open version of our extraction and graph-building pipeline. Transform unstructured text into deterministic, evidence-linked knowledge graphs. Apache 2.0 licensed.
This is a subset of the full TACITUS architecture — designed for experimentation, research, and community building. We believe open source accelerates the entire field.
View on GitHubBuilt With
Neo4j
Graph Database
Next.js
Framework
Google Cloud
Infrastructure
Get notified when new features launch and the API opens to developers.
No spam. Unsubscribe anytime.
Delivered.
Try the public beta, request an enterprise pilot, or build on top of our infrastructure.