Embedded temporal property graph database — bi-temporal facts as a first-class engine primitive.
Kronroe redesigns the embedded graph engine for temporal knowledge evolution. The DuckDB analogy: DuckDB didn't do SQLite better — it redesigned the engine for analytical workloads. Kronroe does the same for graph databases with temporal data.
- Pure Rust — no C dependencies, cross-compiles to iOS/Android/WASM
- Bi-temporal by design — every fact carries valid time + transaction time at the engine level, not the application level
- Embedded — no server required, runs on-device like SQLite
- AI-native — built-in MCP server, vector search, and a high-level agent memory API
| Platform | How | Status |
|---|---|---|
| Native (macOS / Linux / Windows) | cargo add kronroe |
✅ |
| iOS / iPadOS | XCFramework + Swift Package | ✅ |
| WebAssembly (browser) | wasm-pack |
✅ |
| Python | PyO3 bindings — pip install kronroe |
✅ |
| AI agents (MCP) | stdio MCP server — cargo install kronroe-mcp |
✅ |
| Android | UniFFI AAR — planned | 🚧 |
Standard databases record what you know. Bi-temporal databases record what you know and when it was true in the world — independently. This makes Kronroe the right foundation for AI agent memory, audit logs, and any system where facts evolve over time and history must be queryable.
- Graphiti + Neo4j — requires a server; no mobile
- mcp-memory-service — no temporal model at the engine level; no mobile
Website · GitHub · crates.io · PyPI
Licensed AGPL-3.0 · Commercial licences available