Zep Cloud Examples and Integrations: A Repository That Is Not the Product
Zep | Examples, Integrations, & More. It contains example code, framework integrations, and tools for building agent memory with Zep Cloud, Zep's managed agent memory platform.
At a glance
- What is it?
- This repository is a collection of examples, framework integrations, and an ingestion pipeline for Zep Cloud, a managed agent memory platform. It is not the platform itself, and its value depends entirely on whether you plan to use Zep Cloud.
- Who is it for?
- Adopt this repository if you are already a Zep Cloud customer or are evaluating it seriously: the integrations and ingestion tooling are the fastest way to connect frameworks like LangGraph or CrewAI to Zep's managed memory. Do not adopt it if you want a self-hosted memory system; the Community Edition here is deprecated and unsupported.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 2 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 14, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What This Repository Actually Is
The README is explicit: this repository is not Zep's product or service. It is a collection of example code, framework integrations, and tools for building agent memory with Zep Cloud, which is a managed platform. The distinction matters. Many projects on GitHub are the product itself; this one is a supporting cast. The primary language is Python, but the repository covers TypeScript and Go as well. The license is Apache-2.0, which is permissive, but that applies to the code in this repo, not to Zep Cloud itself, which is a paid service. If you are looking for a self-contained memory solution, this is the wrong starting point.
The Structure: Examples, Integrations, Ingestion, and More
The repository is organized into several directories. The examples directory contains apps and snippets in Python, TypeScript, and Go. The integrations directory holds agent-framework integration packages, organized framework-first then language: integrations/<framework>/<language>. Each package is built, tested, and released independently. The ingestion directory contains zep-ingest, a bulk data ingestion pipeline that supports Slack, documents, email, JSON/CSV, and fact triples. There is also an ontology directory with default ontology definitions, a benchmarks directory with memory benchmarks like LoCoMo and LongMemEval, and a zep-eval-harness for evaluation. The legacy directory holds the deprecated Zep Community Edition, which is unsupported. This structure is clear, but it means the repository is a grab bag: you need to know which piece you want before you dive in.
The Ingestion Pipeline: zep-ingest as the Standout Tool
The most substantial tool in this repository is zep-ingest, which has recent releases (v0.3.0 as of late August 2026). It is a bulk data ingestion pipeline that handles multiple source types: Slack, documents, email, JSON/CSV, and fact triples. The README does not provide installation commands or configuration examples, so you cannot tell from this material how to run it beyond knowing it exists. The release cadence suggests active development, but the absence of usage details in the README is a gap. If you need to load historical data into Zep Cloud, this is likely the intended path, but you will need to consult the documentation at help.getzep.com to get started. The benchmarks and eval harness suggest the project cares about measuring memory quality, which is a good sign for anyone evaluating Zep Cloud's actual performance.
Framework Integrations: Breadth with a Catch
The integrations directory lists a wide range of agent frameworks. For Python: Google ADK, Microsoft Agent Framework, Microsoft AutoGen, AG2, CrewAI, LangGraph, LiveKit, Pydantic AI, and Strands Agents. For TypeScript: Google ADK, Mastra, and Vercel AI SDK. For Go: Google ADK. That is a broad spread, and the framework-first organization means you can find the integration for your stack quickly. However, the README warns that each package is built, tested, and released independently. That means not all integrations are equal. Some may be more mature than others, and the release status is not shown in the main README; you have to check integrations/README.md for that. The presence of a CLAUDE.md file for conventions suggests the project uses AI-assisted development, which is worth noting but not a quality judgment.
Getting Running: Commands and Configuration
The README gives only one concrete command: for Python, pip install zep-cloud. For TypeScript, npm install @getzep/zep-cloud. For Go, go get github.com/getzep/zep-go/v3. These are the official SDKs, and they are the entry point to using Zep Cloud. To use any of the examples or integrations, you first need to sign up at www.getzep.com and read the documentation at help.getzep.com. The README does not include any configuration keys, environment variables, or setup steps beyond installing the SDK. That is a limitation of the material: if you want to run a specific integration, you will need to dig into the subdirectory's own README, which is not reproduced here. The lack of a quickstart in the top-level README is a real friction point for a new user.
The Deprecated Community Edition: A Warning
The legacy directory contains the Zep Community Edition, which is deprecated and unsupported. The README links to a blog post announcing a new direction for Zep's open source strategy. This is a significant signal. If you were considering this repository because you wanted an open-source, self-hosted agent memory system, you are out of luck. The Community Edition is dead code. The README points to Graphiti, a separate repository, as the open-source temporal knowledge graph framework that powers Zep. That is the real open-source option, not this repo. So this repository is not a substitute for a self-hosted memory solution; it is a bridge to a cloud service. Anyone who values local control or avoiding vendor lock-in should read that blog post before proceeding.
Alternatives: Graphiti and the Managed Platform Trade-off
The README names Graphiti as the open-source framework that powers Zep. That is the direct alternative to this repository. Graphiti is a temporal knowledge graph framework, which means it takes a different approach: instead of relying on a managed cloud service, you run the graph yourself. This repository, by contrast, is entirely dependent on Zep Cloud. If you want to avoid paying for a managed service or want to keep your data on your own infrastructure, Graphiti is the path. If you are willing to pay for Zep Cloud, then this repository becomes useful as a set of integration examples and ingestion tooling. The trade-off is clear: convenience and managed scaling versus control and cost. The benchmarks in this repo (LoCoMo, LongMemEval) could help you compare Zep Cloud against other memory systems, but you would need to run them yourself to get numbers.
Maintenance and Upgrade Cost
The repository is not archived, and the last push was August 2026, with zep-ingest releases in the same month. That indicates active maintenance. However, the integration packages are released independently, which means you must track each one separately. Upgrades are not centralized; a new version of LangGraph might require a new version of the LangGraph integration, and you will need to check the integrations README for release status. The Apache-2.0 license is permissive, so you can fork and modify the code, but the underlying Zep Cloud service is proprietary. That is a licensing asymmetry: the glue code is open, but the core is not. For any production use, you should verify the license of each integration package you depend on, since the top-level license may not cover every subdirectory. The maintenance cost is low if you use the SDKs directly, but higher if you rely on multiple integrations, because you are at the mercy of each package's release cycle.
Editorial conclusion
Adopt this repository if you are already a Zep Cloud customer or are evaluating it seriously: the integrations and ingestion tooling are the fastest way to connect frameworks like LangGraph or CrewAI to Zep's managed memory. Do not adopt it if you want a self-hosted memory system; the Community Edition here is deprecated and unsupported. Before you start, verify which integration packages are actually released and maintained, check the license of each package (Apache-2.0 applies to this repo, but individual packages may differ), and read the integrations README for release status. If you need an open-source memory framework, look at Graphiti instead, which this README explicitly points to.
Community notes