Skip to main content
For robots only:
Documentation Index: Fetch the complete documentation index at: llms.txt. Use this file to discover all available pages before exploring further.For more information for robots and LLMs, see AI agent integration.
Worlds provides you with an open-source neuro-symbolic infrastructure layer to build agents with persistent, verifiable memory and world models. Traditional memory systems rely on semantic search, which often causes agents to hallucinate or lose structural context. You can bridge this “Reasoning Gap” by fusing hybrid search with the precision of symbolic logic. Worlds ensures your agents navigate reality with deterministic precision.
Worlds API hero image light

​
Overview

Interact with self-contained environments of facts and rules at the network edge.

Bring your own model

Connect your LLM or agent to the Worlds REST SDK.

Construct your world

Ingest data into isolated Worlds where facts live as queryable triples.

Query and reason

Retrieve verifiable logical relationships instead of semantic guesses.

​
Get started

Integrate Worlds into your agentic loops.

Philosophy

Explore the neuro-symbolic and local-first principles driving the platform.

Quickstart

Get up and running with Worlds in minutes.

​
Why neuro-symbolic?

Contrast traditional RAG patterns with the neuro-symbolic architecture.
FeatureSemantic RAGWorlds
Recall typeProbabilistic generationDeterministic precision
LogicImplicit weight-based logicExplicit RDF-based logic
QueryingSimilarity searchSPARQL 1.1 + Hybrid Search
StateEphemeral / StaticMalleable / Verifiable

​
Capabilities

Neuro-symbolic reasoning

Query probabilistic LLM logic and rigid Knowledge Graph truth using SPARQL.

Edge-first architecture

Retrieve localized knowledge with low latency, powered by Deno and Turso.

Malleable knowledge

Edit specific facts within a graph without retraining or re-indexing entire blocks.

Hybrid search

Combine vector similarity, keyword precision, and graph-based relational filters.

Provider agnostic

Connect your preferred models, including OpenAI, Anthropic, Gemini, xAI, and DeepSeek.

Agent-native SDK

Bridge private knowledge using the Tool and MCP-native SDK, complete with ready-to-use functions for schema discovery.

​
Next steps

Integrate Worlds into your agentic loops.

Quickstart

Get up and running with Worlds in minutes.

Academy

Master stateful memory and neuro-symbolic integration.

How Worlds work

Deep dive into items, trust models, and the data pipeline.