---
title: "Zep alternative for temporal memory"
description: "past.dev is a Zep alternative for temporal memory. Compare the temporal model, database, deployment, API surface, and recall output."
canonical: https://past.dev/zep-alternative
last-updated: 2026-08-31
---
# Zep alternative for temporal memory

Source: https://past.dev/zep-alternative

Teams evaluating a Zep alternative can use past.dev as a managed or self-hosted temporal memory API. Zep uses Graphiti, an open-source temporal graph engine that runs on a graph database. Both systems track facts over time. past.dev provides four API endpoints, and recall returns dated evidence with an explicit status.

## When to use Zep

[Zep](https://www.getzep.com/) maintains [Graphiti](https://help.getzep.com/graphiti/getting-started/welcome), an open-source temporal knowledge graph. Graphiti ingests episodes, extracts entities and relationships, records time and marks contradicted relationships as expired. It is suitable for teams that want to operate and customize the engine on their own graph database. See [past.dev vs Zep](/vs/zep) for more detail.

## Why teams evaluate alternatives

- **Graph operations.** Graphiti runs on a dedicated graph database, according to its [documentation](https://help.getzep.com/graphiti/getting-started/welcome). Teams must size, secure and upgrade that database.
- **API scope.** Some teams require only ingestion and recall endpoints instead of direct graph operations.
- **Recall status.** Graph retrieval results require application logic to determine whether evidence is sufficient. past.dev includes that determination in the response status.

## How past.dev compares, design to design

| Property | Zep / Graphiti | past.dev |
| --- | --- | --- |
| Temporal model | Time-stamped facts, expiry on contradiction | Validity windows, supersession links, as-of recall |
| Delivery | Open-source engine + hosted platform | Managed or self-hosted API, four endpoints |
| Storage you operate | Graph database for self-host | None for managed service; Postgres with pgvector for self-hosting |
| Recall output | Search results across graph, text and keywords | Ranked, dated evidence with a status value |
| Unanswerable question | Empty or nearest results | Explicit NoKnownSupport |
| Backfill | Episodes with timestamps | Timestamped ingestion, original dates preserved |

The [benchmarks page](/benchmarks) reports recall, update and abstention results. [How to benchmark agent memory](/benchmarks/methodology) documents the method. We do not publish head-to-head scores without a reproducible configuration.

## When to use each product

- **Choose Zep or Graphiti** when you require an open-source engine, want to customize the graph model and can operate a graph database.
- **Choose past.dev** when you want temporal memory as a managed or self-hosted API with dated evidence and response status. The [quickstart](/docs/memory-api/quickstart) shows the integration.

Both products represent time explicitly. See [temporal knowledge graph](/glossary/temporal-knowledge-graph) and [fact supersession](/glossary/fact-supersession) for the underlying concepts.

## Frequently asked questions

### Is past.dev built on Graphiti?

No. past.dev runs its own temporal graph on Postgres with pgvector. The systems use different engines, storage layers and recall contracts.

### We already run Graphiti. Is there a reason to switch?

A switch may be useful if you want managed operation, explicit response status or recall results measured with a published method. Keep Graphiti if its current deployment and query interface meet your requirements.

## Related

- [past.dev vs Zep, in detail](https://past.dev/vs/zep)
- [mem0 alternative](https://past.dev/mem0-alternative)
- [Temporal knowledge graph](https://past.dev/glossary/temporal-knowledge-graph)
- [How to benchmark agent memory](https://past.dev/benchmarks/methodology)
- [Quickstart](https://past.dev/docs/memory-api/quickstart)