Every platform now sells you an ontology.
In 2026 Palantir, Microsoft, Databricks and Snowflake all launched ontologies. Each one lives inside its own platform: built on a vendor object model, fed mostly from that vendor's tables, and governed by that vendor's permissions. The model of your business becomes the hardest thing to move.
Everything flows in. The ontology, and what you learned building it, stays inside one vendor.
Your ontology is a W3C file you own. Take it, and the knowledge built on it, anywhere.
TrustGraph is the one you own.
Your ontology and the knowledge built on it are standard W3C data, not rows in a vendor's object model. Move them to any RDF store, any cloud or any model, with no lock-in.
Extract. Ground. Govern.
- 1 · Extract
OntologyRAG
Triples that match your ontology, validated against domain, range and subclass constraints.
Read the docs - 2 · Ground
Context Graph
RDF in named graphs, with PROV-O lineage to every source chunk.
Read the docs - 3 · Govern
SHACL-AF Policies
Every query evaluated node by node: Allowed, Filtered or Violation.
Read the docs
Standards you own, not a platform you rent.
| Aspect | TrustGraph | Platform ontologies |
|---|---|---|
| Data model | W3C RDF and OWL | Vendor object model |
| Where it runs | Any cloud, on-prem, self-hosted models | Inside one vendor's data estate |
| Knowledge source | All data: documents, structured data and images | Mostly warehouse tables |
| Access control | SHACL-AF policies on every graph query | Platform permissions |
| Audit trail | PROV-O in queryable named graphs | Platform logs |
| Models | All major LLM APIs, or self-hosted open models | Platform's choice |
| License | Apache 2.0 | Commercial |
W3C from storage to agent.
Every layer is an open standard, so every layer is portable.
Built for teams where decisions carry real risk
Financial services
Counterparties, contracts and obligations extracted against the Financial Industry Business Ontology.
Explore Drug-safety ontologyPharmacovigilance
Adverse-event reports linked to products and evidence, with lineage to every source document.
Explore Threat-intel ontologyThreat intelligence
Threat, event and log intelligence unified in one policy-filtered graph.
Explore W3C SSN/SOSASensor intelligence
Sensors, observations and platforms modeled on the W3C Semantic Sensor Network ontology.
ExploreWhat's new in 2.10
SHACL-AF policy filtering
Allowed, Filtered or Violation for every query, with policy.evaluation audit events.
Guided Graph Traversal
Steer GraphRAG with grounding seeds, graph seeds, language filters and per-hop instructions.
Decision Scoring
A fast, structured decision service: about 35 ms per call.
Cloze Classification
Auto-trains a masked language model from your labeled examples, with confidence scores.
One question, a few classes, value in a day.
Pick one question your agents need to answer. Model the handful of classes it touches, load your data, and see grounded, policy-filtered answers the same day. Grow the ontology as the questions grow.
- Hosted LLMs
- AnthropicOpenAIGoogleMistral
- Open models
- vLLMLM Studiollama.cppOllamaon NVIDIA, AMD or Intel acceleration
- Clouds
- AWSAzureGoogle CloudOVHcloudScalewayand on-prem
- Stores
- Cassandra (graph)Qdrant (vector)
- Deploy
- Docker ComposeKubernetes
$ npx @trustgraph/config
? Platform, stores, LLM provider…
✓ deploy.zip
✓ INSTALLATION.mdRun the wizard, then follow the generated INSTALLATION.md. Prefer a browser? The configuration terminal builds the same bundle. CLI docs
Nothing to installRequest access to the hosted PlaygroundWhat is TrustGraph?
For enterprises deploying AI agents where decisions carry real risk, TrustGraph is the open-source ontology runtime that turns all your data, including documents, structured data and images, into ontology-grounded knowledge, and governs what every agent and user can see from it. Unlike platform ontologies from Palantir, Microsoft, Databricks and Snowflake, TrustGraph is built on W3C standards (RDF, OWL, SHACL, PROV-O) and Apache 2.0, so your ontology, your knowledge and your audit trail stay yours.
How is this different from Palantir's Ontology?
Palantir's Ontology is proprietary and lives inside Foundry. TrustGraph uses W3C standards, is Apache 2.0, and runs anywhere; you can export your ontology and knowledge as Turtle.
Do I need an existing ontology?
No. Start with a handful of classes for one question, or begin with schema-free GraphRAG and add an ontology later.
What is a context graph?
A context graph connects business entities, relationships and evidence so AI agents retrieve knowledge with its full meaning and its source. TrustGraph builds context graphs, the technique Thoughtworks' Technology Radar recommends assessing for agentic AI.
How is this different from AI guardrails?
Guardrails filter what a model says after it has seen the data. TrustGraph's SHACL-AF policies filter what an agent or user can see in the first place: every graph query is evaluated node by node and each result is Allowed, Filtered or a Violation, with the evaluation recorded.
Which standards does TrustGraph use?
W3C RDF (with RDF-star), OWL, SHACL-AF, PROV-O and SPARQL 1.1, plus MCP for agent access. Your ontologies, policies and provenance stay portable with no lock-in.
Is TrustGraph open source?
Yes, the core is Apache 2.0. Enterprise add-ons cover knowledge-based access controls and IAM, an attestation engine and temporal knowledge management, with engineering support for production deployments.
See your ontology govern your agents.
A 30-minute technical walkthrough with the TrustGraph team, built around your data.