agent-memory vs cortexdb
Neo4j Labs' graph-native agent memory: conversations, a POLE+O entity knowledge graph and reasoning traces in one store, with a 16-tool MCP server and hosted or self-hosted backends. — versus — Agent memory plus a knowledge graph with ontology constraints in one SQLite file: pure Go, no server, no embedding model required. Plugins for Claude Code, Codex, dsh and Obsidian.
Both are graph-native agent memory with entities and reasoning traces. Neo4j Labs needs a Neo4j server; CortexDB is one SQLite file.
| agent-memory | cortexdb | |
|---|---|---|
| Stars | 604 | 475 |
| Forks | 110 | 27 |
| Language | Python | Go |
| License | Apache-2.0 | MIT |
| Last activity | 2 days ago | yesterday |
| Topics | memory, knowledge-graphs | memory, knowledge-graphs |
| Curated connections | 10 | 5 |
agent-memory — the curator's take
The pick when memory has to be queryable as a graph instead of a black box: entities resolve and dedupe, reasoning steps get explicit :TOUCHED audit edges to the entities they used, and you can adopt an existing Neo4j graph as long-term memory rather than re-ingesting. Multi-tenant scoping, buffered writes, consolidation primitives and an eval harness are already in the box, and the hosted NAMS tier lets you start with no database to run. Caveats: Neo4j Labs marks it Experimental and community-supported; extraction stacks spaCy/GLiNER/GLiREL plus an LLM pass, so ingest costs real time and tokens; and if you don't want a graph database in the stack at all, a Postgres- or file-backed layer is far less machinery.
cortexdb — the curator's take
Use it when you want local, inspectable memory that is more structured than notes (typed entities, ontology rules, traceable recall) but don't want to run Postgres or Neo4j. Embeds as a Go library, with clients for Rust, Python and Node. Skip it for team-scale shared memory under heavy concurrent writes, or if you just want plain Markdown you can read and edit by hand.