zep-mcp

zep-mcp

MCP server for the Zep Cloud API — a context engineering platform for AI agent memory. Provides tools for memory management, knowledge graph operations, and admin features.

Category
访问服务器

README

zep-mcp

MCP server for the Zep Cloud API — a context engineering platform for AI agent memory. Built with FastMCP and the zep-cloud Python SDK.

📖 The story

I originally built this project back in early 2025 when I was just learning how to code. It worked! ...technically. The old version had a fallback client that returned fake success responses when the API was down 😅, wrapper scripts for every platform imaginable, and enough redundant code to make future-me cringe.

A year later, I've learned a thing or two. So I burned it all down and rewrote it.

  • Before: 2 tangled client layers, 8 tools, fallback mode that lied to you, ~4,000 lines 💀
  • After: 1 clean server entry point, 13 tools covering the full Zep Cloud API, ~770 lines

Keeping the old git history because it's kinda funny lmao

Use cases

I use this for my own content — ingesting video scripts and viewer comments into Zep's knowledge graph to understand what I've covered, what my audience cares about, and what I should make next.

Map your content knowledge graph — ingest your scripts and audience comments. Zep auto-extracts topics, entities (e.g. people, groups, ideas), and the relationships between them across everything you've published.

Full knowledge graph built from content scripts and viewer comments

See what topics cluster together — explore how a specific entity connects to content pieces and topics. Blue nodes are your actual content; the surrounding topics show what themes appear across those videos and what commenters associate with that entity.

Entity relationship view showing startup connected to content and related topics from comments

Find content gaps — query for entities with many topic and entity connections but no linked content nodes. Those are things your audience is already thinking and talking about that you haven't made a video on yet.

Rich Asians entity with many topic connections but no content nodes — a content gap

Setup

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Copy .env.example to .env and fill in your Zep Cloud API key:

cp .env.example .env

Install the pre-commit hook to prevent accidentally committing secrets:

cp scripts/pre-commit .git/hooks/pre-commit

Run the server

source venv/bin/activate
python server.py

Or with FastMCP dev mode:

fastmcp dev server.py

Configuration

Variable Required Description
ZEP_API_KEY Yes Zep Cloud API key
ZEP_TOOLSETS No Comma-separated tool groups to enable. Values: memory, admin. Defaults to memory,admin (all tools).

Tools

13 tools organized by jobs-to-be-done:

Memory tools (memory):

  • add_messages — Add messages to a conversation thread
  • get_context — Get assembled context block for a user
  • add_graph_data — Add data to a knowledge graph (text, JSON, message, or triple)
  • search_graph — Search a knowledge graph for facts, entities, or episodes
  • get_task — Poll status of an async Zep operation

Admin tools (admin):

  • manage_user — User CRUD (create, get, update, delete, list, warm)
  • manage_user_instructions — Manage user summary instructions
  • manage_thread — Thread CRUD (create, get, delete, list)
  • manage_graph — Graph lifecycle (create, get, update, delete, list, clone)
  • manage_graph_structure — Graph schema, ontology, and custom instructions
  • manage_graph_data — Node, edge, and episode inspection and management
  • manage_context_templates — Context template CRUD
  • project_info — Get current Zep Cloud project info

Running tests

source venv/bin/activate
python -m pytest tests/

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选