Back to Gap Map Product / UX / Agent memory & context stores

Graphiti

Zep
open source / Overall score: 3.8

Graphiti builds temporal knowledge graphs for AI agents from conversations and business data, extracting entities and facts with a language model and giving each fact a window of validity, so that outdated facts are invalidated rather than deleted. Applications query the graph with combined semantic, keyword and graph search on Neo4j, FalkorDB or Amazon Neptune. Zep develops it as the open framework inside its hosted service.

Openness

5 high confidence
5.0
license
Apache-2.0 (graphiti-core framework and the reference MCP and REST servers)
source
public
core features withheld
no — the 434-path tree at 9b8a463 carries one root Apache-2.0 LICENSE and no ee/, enterprise/, commercial/ or proprietary/ path; Zep sells its managed context platform, with its own graph database, users, threads and governance, as a separate product built on the framework
commercial
Zep

Graphiti is Apache-2.0, and the whole framework, from extraction and fact invalidation to hybrid search, is in the public repository. Zep sells a managed context platform built on Graphiti, with its own graph database, user and thread management and governance, as a separate product rather than as features cut from the framework.

Adoption

3 medium confidence
3.0

Adoption is counted as monthly PyPI downloads of graphiti-core, the framework's only package. The reference MCP server, which runs from Docker, is not counted.

Capability

5 medium confidence
5.0

Graphiti performs the whole cycle itself: it extracts entities and facts from each episode, merges duplicates, invalidates facts that newer data contradicts while keeping their history, and searches the resulting graph within a namespace. Users, threads and serving are left to the application that embeds it, which is what Zep adds as a hosted service.

Verified 2026-11-20, except some scores, still awaiting re-confirmation