Head to head

ArgosBrain vs Graphiti

Graphiti is the Python library. ArgosBrain is the in-process Rust engine.

What Graphiti does

Domain-general Python library for temporal knowledge graphs over an external graph DB.

What ArgosBrain does differently

Domain-specialized Rust engine, zero external DB, code-aware, sub-ms P99.

What Graphiti is

The baseline, stated fairly.

Graphiti (Apache-2.0; maintained by Zep) is a Python library for real-time temporal knowledge graphs. Not a server — you embed it. Each edge tracks four temporal dimensions (created_at, valid_at, invalid_at, expired_at).

Episodes (unstructured text or structured JSON) are decomposed into entities + labeled relations. Supports learned or prescribed ontology. Backends: Neo4j 5.26+, FalkorDB, Kuzu, Amazon Neptune.

How it actually works

Technical facts.

ArchitectureTemporal graph library over external DB.
Write costLLM calls per episode (entity extraction, relation inference, conflict detection).
Read costHybrid semantic + BM25 + graph traversal; no read-time LLM.
Code-specificNone. Open issue #1300 on temporal decay; built for CRM / compliance / healthcare.

Sources: GitHub · Help docs

Verdict

Where each one wins.

↑ Where ArgosBrain wins

No external DB. Graphiti requires Neo4j/FalkorDB/Kuzu/Neptune. ArgosBrain is petgraph in-process.
Code-native. Graphiti is domain-general; we're domain-specialized.
Rust vs Python — meaningful for sub-ms P99.
Benchmarked on code (LongMemCode). Graphiti isn't.

↑ Where Graphiti wins

Bi-temporal ontology is richer than our current zone + age-category model.
Domain-agnostic — works for CRM, healthcare, compliance, not just code.
Apache-2.0 library is drop-in for Python shops.

When to choose which

Honest recommendation.

Choose Graphiti if

Non-code domain
Python stack
OK with external graph DB

Choose ArgosBrain if

Coding agent
Rust or MCP
No external DB