Playground | Self-Host | Docs | YouTube | Discord | Website
Open Source · Open Standards · Total Transparency
TrustGraph turns enterprise knowledge, systems, and business logic into an ontology-grounded operating graph connected to a policy runtime that governs what AI agents can do.
Agents propose actions. TrustGraph checks them against business meaning, authority, and evidence before execution. Each decision preserves the evidence and policy version behind it as queryable W3C PROV-O provenance.
- Understand your business: Build a connected operating graph of entities, relationships, policies, and evidence from documents and systems, using your own or domain-standard ontologies.
- Govern agent actions: Enforce explicit rules for authority, scope, risk, uncertainty, time, and value at the point of action.
- Verify decisions: Retain the evidence and policy versions needed to explain and reproduce each outcome.
Built on W3C RDF 1.2, OWL, SHACL, and PROV-O, TrustGraph keeps your business knowledge portable. Deploy with Docker or Kubernetes, in the cloud or on-prem, using hosted LLMs or open-weight models with the native TrustGraph inference stack.
To understand why AI struggles in complex use cases, consider Abbott and Costello’s classic "Who's on First?" routine.
Abbott explains the baseball lineup: Who is on first base, What is on second base, and I Don't Know is on third base. Costello is driven mad because he assumes Abbott is asking questions rather than stating the names of the players: Who, What, and I Don't Know.
Two agents cannot communicate if they do not share the same context understanding.
If you feed this scenario into a standard RAG pipeline using vector embeddings or keyword search, it breaks completely.
If a user asks: "Who is playing on first base?"
- The vector database converts the query into an embedding.
- Cosine similarity searches for vectors close to "playing," "first base," and "who."
- Because "Who" is a common pronoun, the embedding space maps it to general inquiries about identity, not the specific name of a baseball player.
- The LLM retrieves irrelevant documents and hallucinates, failing to understand that "Who" is an entity (a Person), not a question.
Cosine similarity operates on fuzzy, statistical probability. It cannot distinguish between the linguistic usage of a word as a pronoun and its usage as a proper noun within a specific, localized context.
An ontology defines the concepts, unique definitions, and relationships that matter in a domain. In the “Who’s on First?” example, it distinguishes a player’s name from their position, so “Who” can be the name of the player at first base without being mistaken for a question. It gives knowledge extraction and retrieval a shared understanding instead of leaving a LLM to infer entity types, definitions, and relationships from each text chunk.
TrustGraph puts that shared understanding to work. Bring an OWL ontology, build your own, or start with a domain standard. For each chunk of source text, TrustGraph retrieves the relevant part of the ontology to guide extraction, turning raw data into more consistent, ontology-enhanced knowledge. Native natural-language retrieval then makes that knowledge available to agents, with its provenance intact.
A semantic intelligence layer built using standards like RDF and OWL, establishes explicit, unambiguous semantics. It doesn't rely on "guessing" based on word proximity; it relies on defined relationships.
Here is the "Who's on First" routine modeled in RDF with an OWL ontology. By structuring data this way, the LLM knows exactly what "Who" means in this context:
@prefix : <http://trustgraph.ai/baseball#> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .
# Ontology Classes
:Player a owl:Class ;
rdfs:subClassOf owl:Thing .
:BaseballPosition a owl:Class .
# Object Properties
:playsPosition a owl:ObjectProperty ;
rdfs:domain :Player ;
rdfs:range :BaseballPosition .
# Data (The Context)
:Who a :Player ;
rdfs:label "Who" .
:What a :Player ;
rdfs:label "What" .
:IDontKnow a :Player ;
rdfs:label "I Don't Know" .
:FirstBase a :BaseballPosition ;
rdfs:label "First Base" .
:SecondBase a :BaseballPosition ;
rdfs:label "Second Base" .
:ThirdBase a :BaseballPosition ;
rdfs:label "Third Base" .
# The Explicit Relationships
:Who :playsPosition :FirstBase .
:What :playsPosition :SecondBase .
:IDontKnow :playsPosition :ThirdBase .When an agent queries the TrustGraph operating graph, it uses SPARQL or GraphRAG to traverse these explicit paths. The agent knows that :Who is a :Player whose :playsPosition is :FirstBase. Hallucination is eliminated because context is structured, not inferred via probability.
- Operating Graph — An RDF 1.2‑compliant named graph system with automated natural language retrieval, semantic filtering, and reranking, so queries return grounded, contextually relevant answers rather than raw search hits.
- Policy Enforcement — Native support for OWL ontologies, enabling formal class hierarchies, property constraints, and logical inference over your knowledge graph.
- Agent Runtime — Bring your own agent framework and integrate via the TrustGraph API Gateway, or use the native TrustGraph Agent Runtime, which traces all agent behavior and links every decision back to its source semantic intelligence with full provenance.
- Knowledge Lifecycle Management — Workspaces, Collections, Flows, and Knowledge Cores give you multiple independent degrees of freedom for isolating, accessing, and versioning semantic knowledge over time.
- Knowledge Interoperability — Built on open standards (RDF 1.2, OWL, PROV-O), TrustGraph stores intelligence in interoperable serializations like Turtle that can be exported or migrated to any RDF-compliant system.
- Unstructured Data Ingest — Converts PDF, DOCX, XLSX, PPTX, HTML, Markdown, CSVs, and images into structured semantic intelligence.
- Full LLM Inference Stack — Connect to all major LLM provider APIs, or self-host open-weight models on Nvidia, AMD, or Intel hardware.
| Dimension | TrustGraph | Conventional Graph Databases (e.g., Neo4j) |
|---|---|---|
| Primary purpose | Operating graph purpose-built for AI: make knowledge unambiguous, traceable, and retrieval-ready for LLMs and agents | General-purpose property graph database for transactional workloads and graph analytics |
| Data model | RDF 1.2 named graphs (quads) with reification — statements are first-class, addressable resources enabling n-ary relationships | Labeled property graph — nodes and edges with key-value properties; no native statement reification |
| Semantic rigor | OWL ontology enforcement: typed entities and properties with formally defined meaning | Schema-optional; semantics live in application code or conventions, not the data model |
| Provenance | Built-in, standards-based (W3C PROV-O); extraction lineage, query traces, and agent behavior stored as queryable graph triples | Not native; provenance must be hand-modeled as ordinary nodes/edges with no standard vocabulary |
| Natural language retrieval | Automated NL-to-graph retrieval with semantic filtering and reranking | Requires manual Cypher queries or add-on vector search with no semantic grounding |
| Agent integration | Native Agent Runtime with full behavioral tracing linked to source intelligence, plus API Gateway for bring-your-own-framework | None; agents access the graph as an external data source with no behavioral traceability |
| Knowledge lifecycle management | Workspaces, Collections, Flows, and Knowledge Cores for isolation, access control, and versioning of semantic intelligence | Database-level separation only; versioning and lifecycle management are application concerns |
| Unstructured data ingest | Integrated pipeline converts PDF, DOCX, XLSX, PPTX, HTML, Markdown, CSV, and images into ontology-typed knowledge | Not included; requires external ETL and custom extraction pipelines |
| LLM stack | Full inference stack: all major provider APIs or self-hosted open-weight models on Nvidia, AMD, or Intel | None; LLM integration is entirely external |
| Interoperability | Open standards throughout (RDF 1.2, OWL, PROV-O); exports to Turtle portable to any RDF-compliant system | Proprietary property graph model; Cypher is not a W3C standard; migration requires data transformation |
| Query paradigm | SPARQL + semantic graph patterns with automated natural language access | Cypher / GQL pattern matching requiring graph expertise |
How many times have you cloned a repo and opened the .env.example to see the dozens of API keys for 3rd party dependencies needed to make the services work? There are only 3 things in TrustGraph that might need an API key:
- 3rd party LLM services like Anthropic, Cohere, Gemini, Mistral, OpenAI, etc.
- 3rd party OCR like Mistral OCR
- The API key you set for the TrustGraph API gateway
Everything else is included.
- Managed Multi-model storage in Cassandra
- Managed Vector embedding storage in Qdrant
- Managed File and Object storage in Garage (S3 compatible)
- Managed High-speed Pub/Sub messaging fabric with Pulsar or RabbitMQ
- Complete LLM inferencing stack for open LLMs with vLLM, TGI, Ollama, LM Studio, and Llamafiles
No need to clone the repo unless you are building from source. TrustGraph deploys as a set of Docker containers. Configure it on the command line in one step:
npx @trustgraph/config
The config process will generate an app config that can be run locally with Docker, Podman, or Minikube. The process will output:
deploy.zipwith either adocker-compose.yamlfile for a Docker/Podman orresources.yamlfor Kubernetes- Deployment instructions as
INSTALLATION.md
For a browser based configuration, try the Configuration Terminal.
The UI provides tools for all major features of TrustGraph. The UI deploys on port 8888 by default.
- Agent Console — Query your agents directly with streaming responses and live explainability event tracking, so you can watch reasoning unfold in real time
- GraphRAG View — Interactive graph RAG queries with a visual explainability DAG and inline provenance display, making it easy to see exactly where answers came from
- Context Explorer — An interactive 3D context graph explorer with dynamic graph loading, BFS neighborhood extraction, edge pulse animation, and multiple navigation views
- Document Ingestion — A complete upload and submission workflow with page and chunk inspection and document structure browsing
- Ontology Workbench — A full ontology editor with class and property trees, OWL/XML and Turtle import/export with round-trip fidelity, circular dependency detection, and safe-delete confirmation dialogs
- Schema Workbench — Interactive schema management with list, create, edit, and delete operations including field and index management
- Prompt Editor — A dedicated prompt editing workflow
There are TS libraries for quick UI integration of TrustGraph services.
| Package | Description |
|---|---|
@trustgraph/trustkit |
Shared component library and design system |
@trustgraph/client |
TypeScript client for TrustGraph APIs |
@trustgraph/react-provider |
React provider for TrustGraph WebSocket connections |
@trustgraph/react-state |
React state management hooks |
@trustgraph/portal |
Portal application showcasing all workflows |
TrustGraph is licensed under Apache 2.0.
Copyright 2024-2026 TrustGraph
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.