How Fluree fits together.
Fluree is the complete context layer for enterprise AI: meaning, entities, policy, lineage and time in one system, between your data and everything that asks questions of it. It runs beside the stack you already have.
Fluree AIThe context layerTrusted by
How your data becomes one governed graph.
Fluree AI resolves, extracts and models your data in one governed graph on FlureeDB. Each capability does one job; together they take data from where it lives to an answer you can trace back to it.
Resolve
Entity resolution.
Lakehouse tables and SQL sources, queried where they live and resolved into golden records with lineage.
ExploreExtract
Document extraction.
PDFs and Office files become entities and relationships, each one kept with the sentence that proves it.
ExploreModel
Context layer management.
Taxonomies and ontologies, reviewed and versioned like a release, then published to every capability.
ExploreFlureeDB
The engine.
The knowledge graph database underneath: policy on every fact, full history and built-in search, in SPARQL, openCypher, GraphQL or JSON-LD.
Explore
Connect where the data lives.
No bulk migration first. Each kind of source has a direct way in, and every interface reads the same governed graph.
Into the graph
- Lakehouse tablesResolve
Iceberg and Parquet, read in place from a REST catalog or straight from object storage. Filters run at the source.
- SQL databases and warehousesResolve
Postgres, MySQL, SQL Server, Oracle, Snowflake, BigQuery and Redshift, through one Trino or Starburst endpoint, or a sidecar for a single database.
- Business appsResolve
CRM, ERP and billing records, resolved into one golden record per customer, product or supplier.
- DocumentsExtract
PDF, Word, PowerPoint, Excel, Markdown and HTML, parsed with structure and position kept, then read against your ontology.
- VocabulariesModel
Taxonomies, glossaries and ontologies from SKOS, OWL, RDF or spreadsheets, versioned and published across Fluree AI.
- Linked dataFlureeDB
JSON-LD, Turtle, TriG and N-Quads load straight into a ledger.
Out to your AI
- AgentsMCP docs
Claude, OpenAI, Gemini, Ollama: any MCP-speaking agent queries the graph as a tool, under the same policy as a person.
- AppsQuery docs
HTTP, SPARQL 1.1, openCypher, GraphQL and JSON-LD on every ledger, or the Rust library inside your own process.
- BI and semantic layersSemantic layer
Keep your metrics layer. Fluree adds the entities, relationships and policy behind it.
- SearchSearch docs
Full-text and vector search inside the engine, joined to graph patterns in the same query.
- Your modelsFluree AI
Point embeddings and extraction at any OpenAI-compatible endpoint, or use managed models on Fluree AI.
Run it hosted, or run it yourself.
One engine and one graph, three ways to run them.
Fluree AI
Serverless and free to start: the full platform with MCP and API access. Pay for the fuel you use.
Start freeFlureeDB
Source-available under BUSL 1.1. A single binary or Docker image, on-prem, air-gapped or in your cloud.
Explore FlureeDBEnterprise
Single-tenant AWS or private deployments, with SSO, SLAs and support.
Talk to salesQuestions architects ask.
The short answers. The docs go deeper.
- Where does a context layer sit in the stack?
- A context layer sits between your data systems and your AI agents, reading from the lakehouse, databases and documents below and answering agents and apps above.
- How does Fluree work with Snowflake?
- Snowflake keeps storage and compute. Fluree queries Iceberg tables where they live through the Snowflake Horizon catalog and adds meaning, entities, policy, lineage and time on top. Agents ask Fluree, and every answer traces back to its source table.
- How does Fluree work with Databricks?
- Databricks keeps storage and compute. Fluree reads Delta tables in place, including through Unity Catalog, and reads Iceberg through Unity’s Iceberg REST catalog. Delta tables stay where they are; only Iceberg tables can be copied into Fluree.
- Does Fluree replace our lakehouse or catalog?
- Your lakehouse keeps storage and compute. Fluree is the complete context layer on top: meaning, entities, policy, lineage and time in one system. Already run a catalog? Fluree reads from it. You don’t need one to start.
- How do AI agents connect?
- Over MCP. Claude, OpenAI, Gemini, Ollama and any other MCP-speaking agent can query the graph as a tool, under the same policy as a person. Apps use HTTP, SPARQL, openCypher, GraphQL or JSON-LD.
- Where is access policy enforced?
- Inside the query engine, on every read. Policies are stored as data beside the facts they govern, so a fact a person or agent may not see is never returned.
- Do we have to copy our data into Fluree first?
- No. Iceberg tables and SQL sources are queried where they live, through a W3C R2RML mapping. When you want a native copy of a lake table, Fluree builds a verified twin and keeps it in sync.
- Is FlureeDB open source?
- No. FlureeDB is source-available under the Business Source License 1.1: free in production, including inside your own product; a commercial license is needed only to offer it as a hosted database service. Each release converts to Apache 2.0 on its fourth anniversary. The full source is on GitHub.
- Do we have to adopt all of Fluree AI at once?
- No. Start where it fits: golden records from your structured data, graph facts from your documents, governed vocabularies, or FlureeDB on its own. It’s one context layer on one graph, so each part you add builds on the last.
See how it fits your stack.
Bring one question your current stack can’t answer. We’ll show you the graph that answers it, with receipts.
