How Causantic compares to existing AI memory systems, and why it takes a different approach.
| System | Local-First | Temporal Decay | Graph Structure | Self-Benchmarking | Hop-Based Distance |
|---|---|---|---|---|---|
| Causantic | Yes | Hop-based | Causal graph | Yes | Yes |
| Mem0 | No (Cloud) | None | Paid add-on | No | No |
| Cognee | Self-hostable | None | Triplet extraction | No | No |
| Letta/MemGPT | Self-hostable | Summarization | None | No | No |
| Zep | Enterprise | Bi-temporal | Temporal KG | No | No |
| Supermemory | Cloudflare | Dual timestamps | Secondary | No | No |
| A-MEM | Research only | None | Zettelkasten | No | No |
| GraphRAG | Self-hostable | Static corpus | Hierarchical | No | No |
Cloud API with two-phase extraction/update pipeline. LLM extracts facts, then decides ADD/UPDATE/DELETE/NOOP against existing memories. Triple-store hybrid (Vector + Graph + KV). Graph memory is a paid add-on. No temporal decay — memories mutated in place. 66.9% on LOCOMO benchmark.
ECL pipeline (Extract-Cognify-Load) with LLM-based triplet extraction. 12 search modes including graph completion and Cypher queries. Incremental loading (unlike GraphRAG). 100% LLM-dependent — no traditional NLP fallback. Scalability issues (1GB takes ~40 min). 92.5% accuracy.
OS-inspired virtual memory with Main Context (RAM) and External Context (Disk). Agent manages its own memory via tool calls. Recursive summarization is lossy. No graph structure. 93.4% on Deep Memory Retrieval.
Temporal Knowledge Graph via Graphiti engine with bi-temporal model. Best-in-class temporal reasoning among production systems. Enterprise/cloud-focused with higher latency (1.29s p50). 94.8% on DMR — highest among production systems.
Zettelkasten-inspired agentic memory with bidirectional linking. Only system with true associative memory evolution. Doubles performance on multi-hop reasoning. Research paper, not production-ready.
| Gap | Current Landscape | Causantic's Approach |
|---|---|---|
| Local-first + sophisticated | Cloud systems are sophisticated; local systems are simplistic | Full causal graph + clustering + hybrid search, all on your machine |
| Hop-based decay | Wall-clock time or none | Logical D-T-D hops preserve cross-session continuity |
| Direction-specific retrieval | Symmetric or none | Backward (dies@10 hops) vs forward (delayed, dies@20 hops) |
| Self-benchmarking | No system measures its own retrieval quality | Built-in benchmark suite with tuning recommendations |
| Claude Code native | General-purpose or platform-agnostic | Purpose-built hooks, MCP tools, and CLAUDE.md generation |
Most memory systems optimize for storing memories. Causantic optimizes for retrieving the right context at the right time — using hybrid BM25+vector search with causal chain walking for episodic narrative context. (The 4.65× augmentation figure was a v0.2 research result using sum-product traversal, since replaced by chain walking — see experiments/graph-traversal.md.)
Condensed from the full feasibility study. See the archive for detailed per-system analysis including architecture diagrams and benchmark methodology.