Synapse is a high-performance, purely symbolic knowledge graph system designed to serve as the long-term, determinable memory for AI agents.

Built entirely in Rust, Synapse relies on a formal logical reasoning core instead of probabilistic vectors. This guarantees absolute certainty in data retrieval, provenance tracking, and deductive reasoning capabilities.

• Blazing Fast Core: Powered by Rust and Oxigraph for low-latency, strictly symbolic graph operations.
• Markdown-Graph Symbiosis: Bidirectionally syncs human-readable Markdown files (via YAML frontmatter and WikiLinks) with the active RDF graph.
• Reasoning Engine: Built-in, in-memory OWL-RL and RDFS fixed-point reasoning strategies to derive implicit knowledge safely.
• Scenario Marketplace: Dynamic loading of domain-specific "scenarios" (ontologies + seed data) to instantly equip agents with specialized schemas.
• Native MCP & UI Extensions: Seamlessly integrates as a Model Context Protocol server, supporting dual-transport (HTTP/SSE or Standard I/O) and rich Ext-App HTML/d3 UI rendering.
• Ontology-Driven: Automatically loads standard ontologies (Schema.org, PROV-O, SKOS, etc.) via the core scenario.

Ensure you have Rust installed, then build the release binary:

git clone https://github.com/pmaojo/synapse-engine.git
cd synapse-engine
./start_rust_server.sh

Add Synapse to your MCP client configuration to enable direct LLM access to the symbolic knowledge graph:

"mcpServers": {
  "synapse": {
    "command": "/path/to/synapse-engine/target/release/synapse",
    "args": ["--stdio"],
    "env": { 
      "GRAPH_STORAGE_PATH": "/path/to/synapse-engine/data/graphs"
    }
  }
}

• sparql_query: Execute strict, graph-based SPARQL queries to traverse deterministic memory.
• index_markdown_directory: Recursively map a folder of Markdown files into the RDF graph.
• get_entity_neighborhood: Extract the immediate surrounding entities (1-hop) for contextual retrieval.

• ui://synapse/dashboard: Returns an interactive HTML dashboard view of the engine metrics.
• ui://synapse/graph/{uri}: Returns a D3.js powered subgraph visualization for the requested entity.

Synapse enforces structural integrity via Scenarios. A Scenario bundles schemas and data:

1. Ontologies: Formal schema definitions (OWL/TTL) that define classes and property chains.
2. Seed Data: Initial foundational knowledge triples.

• Core: Essential ontologies (Schema.org, PROV-O, SKOS, FOAF, Memory) loaded automatically at startup.
• Research Assistant: Specialized ontologies designed for tracking academic papers and citations.

All probabilistic ML (embeddings, vectors, fastembed) has been explicitly purged. The system relies 100% on explicit RDF triples, ensuring zero hallucinations in memory recall.

The engine implements a multi-strategy materialization loop:

• RDFS: Efficient class (rdfs:subClassOf) and property (rdfs:subPropertyOf) transitivity.
• OWL-RL: Deterministic fixed-point inference for axioms like owl:SymmetricProperty, owl:TransitiveProperty, and owl:inverseOf.

Every fact ingested into the graph is stored within an isolated Named Graph, linked via PROV-O properties (prov:wasDerivedFrom, prov:generatedAtTime). You always know exactly where and when an agent learned a piece of information.

We welcome contributions! Please see CONTRIBUTING.md for guidelines on how to get started.

This project is licensed under the MIT License - see the LICENSE file for details.

Developed by the Synapse Team