Utopia: Open-Source Enterprise World Model Goes Beyond RAG with Bitemporal Graphs

Utopia is an open-source project (GitHub: ~8.0k stars, 1.1k forks as of September 30, 2026, developed by DeepLethe) that positions itself as an Enterprise World Model . Unlike traditional RAG, which retrieves the latest document snippets, Utopia builds a knowledge layer that preserves the evolution of business facts over time.

1. RAG Finds Documents, But Historical Facts Need More Than the Latest Version

RAG answers questions like "What is the refund policy?" by fetching current documents. However, enterprise decisions often require knowing what was true at a specific past moment — e.g., which payment terms applied to a contract on March 15, 2024, after multiple amendments. Utopia's README illustrates this with a Master Services Agreement example: the system must return the effective payment terms for that date and cite the exact amendment document.

If the knowledge layer only stores the latest value (e.g., customer_type updated from SMB to Enterprise in April), an agent reviewing a February pricing decision would incorrectly apply the April status. Utopia addresses this by embedding time into the knowledge base so agents can ask: "What facts did the system believe were true at that time?"

2. Bitemporal Graph: Two Timelines for Every Fact

Utopia's core data structure is a Bitemporal Graph . Each fact carries two independent timelines:

Valid time — when the fact was true in the real world (e.g., contract effective March 1).

Transaction time — when the system recorded or learned the fact (e.g., ERP synced on March 5).

When a correction occurs (e.g., contract amount entered wrong in April, fixed in May), Utopia does not overwrite the old record. Instead, it ends the previous fact's validity interval and creates a new version with a link to its predecessor. This preserves the full evolution of business state, enabling queries like "What amount did the system show on March 10?"

3. Conflict Resolution, Entity Resolution, and Human Review Are First-Class Citizens

Enterprise data arrives from CRM, ERP, contracts, emails, etc., often describing the same entity inconsistently. Utopia integrates:

Ontology — defines entity types, relationships, cardinality constraints.

Entity Resolution — matches by name/alias, then embedding similarity , then falls back to LLM judgment.

Conflict Detection — flags low-confidence extractions, duplicate entities, and cardinality violations.

Review Queue — routes uncertain cases to humans.

Conflicts are not automatically resolved by "newest wins." Options include closing the old fact and accepting the new, keeping both versions, or rejecting the new fact. Ontology constraint conflicts let users retract facts, modify rules, or accept the conflict. Derived facts (via forward-chaining rules like transitive/symmetric/inverse relations) are generated with derivation disabled by default because erroneous rules propagate errors. Derived facts are tagged separately; explicit facts always override derived ones .

4. MCP Integration: Agents Query Typed, Dated State; Writes Are Proposals

Utopia exposes an MCP Server so MCP-compatible clients (Claude Desktop, Cursor) can query the knowledge base with permissioned tokens. The design principle: "Reads are typed and dated, writes are proposals." Agents receive pre-organized business state — entity identity, fact type, validity interval, provenance, replacement links — not raw text chunks. Write attempts enter as Proposals requiring human confirmation.

A Decision Ledger (append-only) records every confirmation, rejection, merge, split, or graph rebuild with actor, timestamp, target object, and action. History is never overwritten by current state changes. This becomes critical when agents participate in contracts, procurement, or risk control and auditors ask: "Why was this fact considered true? What was the source? Who approved the change? Was the state at decision time the same as today?"

5. "Enterprise World Model" Is a Direction, Not a Mature Category

The term evokes RL world models (e.g., Ha & Schmidhuber, 2018), but Utopia does not simulate enterprise dynamics or predict future states. It builds a queryable, updatable, versioned, provenance-tracked business state layer — customers, contracts, organizations, relationships, fact validity, system knowledge time, and modification history.

Roadmap items (Decision Intelligence, decision replay, Execution Gate, Business Rules, Agent Memory over MCP) remain in development. The project is at v0.1 ; the schema may change, migrations are forward-only with no rollback, and production use requires version pinning and backups.

The shift is significant: enterprise context for agents is moving from a static document set for model consumption to a continuously maintained, versioned business state with built-in conflict handling, provenance, and auditability. Whether Utopia can stay accurate and maintainable without turning ontology, review, and version management into new engineering burdens is the key question beyond v0.1.