TrustGraph

Overview

TrustGraph is an open-source context graph platform with a deterministic agent runtime. It turns enterprise knowledge, systems, and business logic into an ontology-grounded operating graph, then checks every action an AI agent proposes against business meaning, authority, and evidence before execution. Decisions can be allow, pause, or deny, and each decision records the evidence and policy version behind it as W3C PROV-O provenance.

Key Features

• Ontology-grounded operating graph built from documents and systems using OWL, RDF 1.2, SHACL, and domain or custom ontologies.
• Deterministic agent runtime that checks proposed actions against semantic policy for authority, scope, risk, uncertainty, time, and value.
• Allow, pause, or deny decisions recorded with evidence and policy version as queryable W3C PROV-O provenance.
• Deployment with Docker Compose or Kubernetes, on cloud or on-prem, with hosted LLM APIs or self-hosted open-weight models.
• Unstructured data ingest for PDF, DOCX, XLSX, PPTX, HTML, Markdown, CSV, and images into ontology-typed knowledge.
• UI tools including Agent Console, GraphRAG View, Context Explorer, Document Ingestion, Ontology Workbench, Schema Workbench, and Prompt Editor.

Use Cases

• Finance and banking: enforce payments, approvals, and delegated authority on every agent action.
• Agentic threat intelligence: unify threat, event, and log intelligence in one governed graph.
• Pharmacovigilance: support traceable drug-safety workflows with evidence behind every decision.
• Aerospace, defense, and telecom: high-assurance agents where decisions must be auditable and reproducible.

Getting Started

• Run npx @trustgraph/config to generate deployment configuration.
• The config process outputs deploy.zip with docker-compose.yaml for Docker or Podman or resources.yaml for Kubernetes, plus INSTALLATION.md.
• A browser-based Configuration Terminal is available for configuration.
• Documentation includes Getting Started Guides, Developer APIs and CLI, and Deployment Guides.

Deployment & Requirements

• Deploy with Docker Compose or Kubernetes.
• Can run on every major cloud or on-prem.
• Supports hosted LLM APIs or open-weight models on your own accelerators.
• Includes managed multi-model storage in Cassandra, vector embedding storage in Qdrant, file and object storage in Garage, and high-speed pub/sub messaging with Pulsar or RabbitMQ.
• Native inference stack supports vLLM, TGI, Ollama, LM Studio, and Llamafiles.

Before You Adopt

• License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
• The core is open source, while some enterprise capabilities are provided as add-ons with production engineering support.
• Using third-party LLM providers may require API keys.
• Using third-party OCR services may require API keys.