changedetection-mcp
An MCP server that gives AI agents native access to ChangeDetection.io, enabling watch management, change history, and tag organization for website change detection.
README
ChangeDetection.io MCP Server
<p align="center"> <a href="https://buymeacoffee.com/rusty4" target="_blank"> <img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" height="50"> </a> </p>
<!-- mcp-name: io.github.rusty4444/changedetection-mcp -->
An MCP (Model Context Protocol) server that gives AI agents native access to ChangeDetection.io — the open-source website change detection and monitoring platform.
Features
- Watch Management — Create, list, update, delete, and trigger rechecks on watches
- Change History — Browse snapshot history and get diffs between any two points in time
- Per-watch action fuse —
get_snapshot_diffarms a configurable limit on follow-up mutating actions for that watch so one noisy page cannot cascade into unlimited rechecks or edits - Tag Management — Organise watches with tags/groups
- Search — Full-text search through watches by URL or title
- System Info — Quick health and stats readout
Why This Exists
No MCP server existed for ChangeDetection.io despite it being one of the most popular self-hosted tools (31k+ GitHub stars). This server fills that gap, letting AI agents:
- "Watch this product page and notify me when it changes"
- "Show me all my failed watches"
- "What changed on this page yesterday?"
- "Set up a watch for this job listing"
Quick Start
Prerequisites
- A running ChangeDetection.io instance (self-hosted or SaaS)
- An API key from Settings → API
Installation
pip install changedetection-mcp
Configuration
Add to your MCP client config (Claude Desktop, Cursor, Hermes Agent, etc.):
{
"mcpServers": {
"changedetection": {
"command": "changedetection-mcp",
"env": {
"CHANGEDETECTION_BASE_URL": "http://localhost:5000",
"CHANGEDETECTION_API_KEY": "your-api-key-here"
}
}
}
}
Usage with Hermes Agent
Create ~/.hermes/skills/changedetection/SKILL.md (see skill/ directory) or configure as an MCP server:
hermes mcp add changedetection --command "changedetection-mcp"
hermes config set changedetection.base_url "http://localhost:5000"
hermes config set changedetection.api_key "your-api-key"
API Coverage
| Tool | Description |
|---|---|
list_watches |
List all watches (optional tag filter, limit/offset) |
get_watch |
Get full details of a single watch |
create_watch |
Create a new watch from a URL |
update_watch |
Update an existing watch (pause, rename, change schedule, etc.) |
delete_watch |
Delete a watch and all its history |
recheck_watch |
Trigger an immediate recheck |
get_watch_history |
List change history snapshots for a watch |
get_snapshot_diff |
Get diff between two snapshots |
search_watches |
Full-text search by URL or title |
list_tags |
List all tags/groups |
create_tag |
Create a tag for organisation |
get_system_info |
Get server stats (watch count, uptime, version) |
Safety Fuse: Per-Watch Action Limit
When an agent asks for a page diff, that diff can prompt follow-up actions such as immediate rechecks, watch edits, or deletes. To stop one noisy site from triggering a cascade, get_snapshot_diff arms a per-watch fuse before returning.
By default, each diff allows 3 mutating MCP actions for that watch. The fuse applies to:
recheck_watchupdate_watchdelete_watch
After the budget is exhausted, those tools return a clear blocked message instead of calling the ChangeDetection.io API. The fuse is in-process and per watch UUID; watches without an armed fuse are unaffected.
Configuration:
# Default: 3. Set 0 to disable the fuse globally.
export CHANGEDETECTION_MCP_ACTION_LIMIT_PER_WATCH=3
Per call, override the default with the optional action_limit argument on get_snapshot_diff:
get_snapshot_diff(uuid="...", from_timestamp="previous", to_timestamp="latest", action_limit=1)
Use action_limit=0 to disable the fuse for that watch, or call get_snapshot_diff again with a higher value to re-arm it.
Development
git clone https://github.com/rusty4444/changedetection-mcp.git
cd changedetection-mcp
uv venv
source .venv/bin/activate
uv pip install -e ".[dev]"
License
MIT — see LICENSE.
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。