World Wide Knowledge Graph
WWKG, the World Wide Knowledge Graph, is peer-to-peer knowledge graph infrastructure for AI agents and data products, with SPARQL, Cypher, and GQL in one engine. More than a database—it’s an intelligent layer over content-addressed storage where relationships are first-class citizens, every commit is immutable, and only you hold the keys. Branch, merge, and time-travel your data like code.
The Problem
Centralized servers with mutable state. No versioning. No encryption. No way to verify that the data you received is the data that was stored. And vendors are chasing AI platforms and GraphRAG suites instead of fixing the foundation. The market is split between semantic graphs (SPARQL, RDF) and property graphs (Cypher, GQL)—both right for different jobs, but once you pick a side you can’t move, there’s no credible unification, and you can’t cherry-pick the best of both. Whether you use SPARQL, Cypher, or GQL—knowledge graphs deserve better infrastructure.
Centralized servers
Single point of failure and control
No integrity proofs
Mutable state with no way to detect tampering
No versioning
Can't branch, merge, or compare changes
No provenance
Can't prove who changed what, when, or why
No temporal queries
History is discarded — no way to query the past
No encryption
Data visible to operators and cloud providers
Links that rot
Move a server and references break silently
No native P2P
Federation bolted on as afterthought
No offline mode
Requires constant server connection to function
Single-machine ceiling
Indexes and storage bound to one server
No LPG reasoning
Cypher and GQL users get zero ontology support
Bolt-on vector search
Embeddings stored as node properties — no versioning, no encryption, no reasoning integration
Features
Version every change. Prove nothing was tampered with. Collaborate without a central server. Run offline. Encrypt everything—and still query it.
Every resource gets a cryptographic hash. If data changes, the hash changes. You can prove what was stored, when, and that nothing was altered.
Peer-to-peer distribution built in. Work offline, sync when ready. Content is served from the nearest node—no single point of failure, no vendor lock-in.
Work on draft changes in isolation, then merge. Query any point in history. Compare versions. Full audit trail from day one.
Think WhatsApp groups for data. You decide who joins the workspace, and everything inside is end-to-end encrypted. Storage providers, network peers, and cloud operators never see plaintext. Only members hold the keys.
The graph derives its own embeddings — no external model needed. Vector similarity composes with ontology reasoning in a single query. Encrypted, versioned, and distributed.
The strengths of each paradigm flow both ways. Cypher and GQL gain ontology reasoning and validation—a first for property graphs. SPARQL gains the traversal patterns developers expect.
Data and indexes are content-addressed blocks that distribute across the network. No node needs all the data. Scale grows with the number of peers, not the size of one server.
Content-addressed links point to what the data is, not where it lives. Data can move between nodes without breaking references. Same content, same hash, forever.
One unified model ends the RDF/property graph divide. SPARQL, Cypher, and GQL compile to a single optimizer—no translation layer, no dual engine. The strengths of each paradigm available to every user.
Identical content is stored once, no matter how many nodes publish it. Saves storage, saves bandwidth, and makes syncing between peers fast.
PDFs, images, and video as first-class graph values. Content-addressed, encrypted, and queryable. No external file stores, no broken links.
AI Agents
Most AI agent frameworks give their agents a search box over embedded documents. WWKG gives them a structured world model—with built-in guardrails that keep humans in charge of what reaches production.
Each agent gets a workspace scoped to its domain—only the data it needs, nothing more. The workspace is the agent’s world model: a structured, queryable, encrypted sandbox.
Agents query real relationships, not text fragments. Graph-derived embeddings add similarity search over structured knowledge. Reasoning, traversal, and vector search compose in a single query.
Agents write to branches—never to main. They can create, modify, and experiment freely. A human reviews the result and decides what gets merged. The agent cannot override this.
When agents consume and produce knowledge, that has cost and value. WWKG’s data economy makes both visible—what data was accessed, what was created, and what it’s worth.
Trust Without Blockchain
Organizations evaluating blockchain for data integrity, provenance, or decentralization often discover the overhead doesn’t justify the outcome. WWKG delivers the trust properties that make blockchain compelling—without requiring any of it.
Every piece of data carries its own cryptographic proof. Modification is not just detectable—it’s impossible to hide.
Trust comes from cryptography, not blockchain consensus. No gas fees, no validators, no staking requirement, no consensus delays.
Every change is signed and recorded. Who changed what, when, and why—queryable from day one.
Peers exchange data directly. No central server to compromise. No blockchain to maintain.
Fetch data from any peer and prove it’s authentic. The data itself is the proof.
No block confirmation waits. Sub-millisecond query performance with the integrity guarantees of a distributed ledger.
Technology
Built in Rust for performance and safety. No garbage collector. No runtime exceptions. Zero-copy block access with compile-time memory safety.
Comparison
How does WWKG compare to established graph databases and the new generation of distributed knowledge graph platforms?
Dive into the design documents and see how WWKG rethinks every layer of the knowledge graph stack.