apache-atlas-mcp
Connects LLM agents to Apache Atlas for searching, tracing lineage, and managing metadata. Read-only by default, with optional write mode for creating entities and classifications.
README
<p align="center"> <img src="https://raw.githubusercontent.com/DanMeon/apache-atlas-mcp/main/assets/banner.svg" alt="Apache Atlas MCP Server" width="900" /> </p>
<p align="center"> <a href="https://github.com/DanMeon/apache-atlas-mcp/actions/workflows/ci.yml"><img src="https://github.com/DanMeon/apache-atlas-mcp/actions/workflows/ci.yml/badge.svg" alt="CI" /></a> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT" /></a> <a href="https://www.python.org/downloads/"><img src="https://img.shields.io/badge/python-3.11+-blue.svg" alt="Python 3.11+" /></a> </p>
A community-built MCP (Model Context Protocol) server that connects LLM agents to Apache Atlas metadata governance platform. Not affiliated with the Apache Software Foundation.
Read-only by default. Write operations (create/delete entities, manage classifications) require explicit opt-in via --write flag or ATLAS_ALLOW_WRITE=true.
Why?
Apache Atlas is a widely-used open-source metadata governance framework for Hadoop ecosystems and beyond. With the 2.4.0 release (Jan 2025) breaking a 2-year gap and 2.5.0 (RC) adding PostgreSQL backend support and Trino extractor, Atlas is seeing renewed activity.
This MCP server lets AI agents:
- Search data assets across your entire metadata catalog
- Trace lineage — understand how data flows from source to destination
- Browse glossaries — look up business terms and definitions
- Inspect types — understand the metadata schema structure
- Manage classifications — apply governance tags like PII, GDPR, Confidential (write mode)
- Create/delete entities — modify metadata catalog (write mode)
<p align="center"> <img src="https://raw.githubusercontent.com/DanMeon/apache-atlas-mcp/main/assets/search-demo.png" alt="Search entities demo" width="800" /> </p>
Quick Start
Installation
# Using uv (recommended)
uv add apache-atlas-mcp
# Using pip
pip install apache-atlas-mcp
# Run without installing
uvx apache-atlas-mcp
Configuration
Set environment variables to connect to your Atlas instance:
export ATLAS_BASE_URL=http://your-atlas-server:21000
export ATLAS_USERNAME=your-username
export ATLAS_PASSWORD=your-password
All configuration options:
| Variable | Default | Description |
|---|---|---|
ATLAS_BASE_URL |
(required) | Atlas server URL (e.g., http://localhost:21000) |
ATLAS_USERNAME |
(required) | Authentication username |
ATLAS_PASSWORD |
(required) | Authentication password |
ATLAS_VERIFY_SSL |
true |
Verify SSL certificates |
ATLAS_TIMEOUT |
30 |
HTTP request timeout (seconds) |
ATLAS_ALLOW_WRITE |
false |
Enable write operations (create, delete, classify) |
Usage with Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"apache-atlas": {
"command": "uvx",
"args": ["apache-atlas-mcp"],
"env": {
"ATLAS_BASE_URL": "http://your-atlas-server:21000",
"ATLAS_USERNAME": "your-username",
"ATLAS_PASSWORD": "your-password"
}
}
}
}
To enable write operations:
{
"mcpServers": {
"apache-atlas": {
"command": "uvx",
"args": ["apache-atlas-mcp", "--write"],
"env": {
"ATLAS_BASE_URL": "http://your-atlas-server:21000",
"ATLAS_USERNAME": "your-username",
"ATLAS_PASSWORD": "your-password"
}
}
}
}
Usage with Claude Code
# Read-only (default)
claude mcp add atlas-server -- uvx apache-atlas-mcp
# With write operations enabled
claude mcp add atlas-server -- uvx apache-atlas-mcp --write
Run Directly
# Read-only (default)
apache-atlas-mcp
# With write operations enabled
apache-atlas-mcp --write
Available Tools
Read-Only Tools (always available)
Entity Search
| Tool | Description |
|---|---|
search_entities |
Basic search by keyword, type, or classification |
dsl_search |
Advanced search using Atlas DSL query language |
quick_search |
Fast partial-match search (autocomplete-style) |
get_entity |
Get full entity details by GUID |
get_entity_by_attribute |
Get entity by unique attribute (e.g., qualifiedName) |
get_entities_bulk |
Fetch multiple entities by GUIDs |
Lineage
| Tool | Description |
|---|---|
get_lineage |
Trace upstream/downstream data flow by GUID |
get_lineage_by_attribute |
Trace lineage by unique attribute |
Classifications
| Tool | Description |
|---|---|
get_classifications |
List all classifications on an entity |
Glossary
| Tool | Description |
|---|---|
list_glossaries |
List all business glossaries |
get_glossary |
Get glossary details with terms and categories |
get_glossary_terms |
List terms in a glossary |
get_glossary_term |
Get full details of a glossary term |
Type Definitions
| Tool | Description |
|---|---|
get_all_type_definitions |
Get the complete Atlas type system |
get_type_definition |
Get a specific type definition by name |
get_entity_type_definition |
Get an entity type with all attributes |
Write Tools (requires --write or ATLAS_ALLOW_WRITE=true)
| Tool | Description |
|---|---|
create_entity |
Create or update an entity |
delete_entity |
Soft-delete an entity |
add_classification |
Apply a classification tag (with lineage propagation) |
remove_classification |
Remove a classification from an entity |
Architecture
LLM Agent <--> MCP Protocol <--> Apache Atlas MCP Server <--> Atlas REST API v2
|
FastMCP + httpx
|
Apache Atlas Instance
(HBase / PostgreSQL backend)
The server wraps the Atlas REST API v2 using:
- FastMCP for MCP protocol handling
- httpx for async HTTP communication
- Pydantic V2 for data validation and serialization
The REST API is backend-independent — it works regardless of whether Atlas uses HBase, Cassandra, or the new PostgreSQL backend (Atlas 2.5.0+).
Compatibility
- Apache Atlas: 2.1.0+ (tested with 2.3.2 and 2.4.0)
- Python: 3.11+
- Authentication: HTTP Basic (Atlas file-based auth, LDAP, or AD — all use Basic HTTP headers)
Development
# Clone the repository
git clone https://github.com/DanMeon/apache-atlas-mcp.git
cd apache-atlas-mcp
# Install dependencies
uv sync --group dev
# Run tests
uv run pytest tests/ -v --cov
# Lint & type check
uv run ruff check src/ tests/
uv run pyright src/apache_atlas_mcp/
License
MIT
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。