Just Publish MCP Server
Publish the website you built with AI to a live public URL — straight from chat, no setup. Enables deploying static sites and updating them with edit tokens.
README
Just Publish — MCP Server
Publish the website you built with AI to a live public URL — straight from chat, no setup.
Describe a site to your AI assistant — ChatGPT, Claude, Gemini, or any MCP-enabled client — and it publishes the site and hands back a live URL. No git, no CLI, no build step, no dashboard, and no login to fumble through. You hand it your files, it hands back a working link. Built for the person who made a site with AI and just wants it online.
- Website: https://justpublish.ai
- MCP endpoint:
https://mcp.justpublish.ai/ - Registry name:
ai.justpublish/just-publish - Transport: Streamable HTTP (remote — nothing to install)
How it works
Call deploy with your files (HTML, CSS, JS, images) and an email. You get back a
live url, a site_id, and an edit_token. Keep the token and you can update the
same site anytime — either a full redeploy or a single-file edit. That's the whole
flow.
Connecting
{
"mcpServers": {
"just-publish": {
"type": "streamable-http",
"url": "https://mcp.justpublish.ai/"
}
}
}
Then ask your assistant to publish a site. It returns the public URL.
Tools
| Tool | What it does |
|---|---|
deploy |
Publish a new static site (HTML/CSS/JS/images), or fully replace an existing one. Returns the live URL, a site_id, and an edit_token. |
get_site_files |
Read a published site's files so the assistant can make a targeted edit. Requires site_id + edit_token. |
update_site_file |
Change one or a few files in place (merge). Requires site_id + edit_token. |
Resources
Static markdown guidance the client can load into context before it picks a tool:
| Resource | Purpose |
|---|---|
just-publish-overview |
What Just Publish is and which tool to use when. |
just-publish-edit-lifecycle |
The read → edit → update loop. |
just-publish-file-conventions |
Paths, encoding, size limits, routing. |
just-publish-invariants |
index.html required; deploy replaces vs. update_site_file merges; edit_token is the only auth. |
Plus a UI resource for the publish-result widget.
Agent Skill (ChatGPT / Codex / any SKILL.md runtime)
This repo also hosts the just-publish Agent Skill — a SKILL.md-format
package that teaches an agent the full publish flow (including saving the
edit_token and the email-verification step) before it ever calls a tool.
To install in Codex / ChatGPT, ask the assistant to install the skill from this tree:
https://github.com/just-done/just-publish-mcp/tree/main/skills/just-publish
(Codex's skill-installer takes a GitHub tree URL.) For a manual install, copy
skills/just-publish/ into $CODEX_HOME/skills/ (default ~/.codex/skills/)
or your runtime's equivalent (e.g. ~/.claude/skills/).
The same package is served for discovery scanners and other runtimes at justpublish.ai/.well-known/agent-skills/index.json (Agent Skills Discovery RFC, sha256-digested). The canonical source lives in the Just Publish server codebase; this tree is its published mirror and is kept in lockstep — the two are compared on every server deploy.
Who it's for
Non-technical builders and AI agents that generate a static site and need it hosted in one step. If you can describe a page, you can publish it.
Good to know
- No accounts required. Edit access is held by the
edit_tokenreturned at publish time — save it to update the site later. - Static sites only — files in, URL out. No frameworks or build pipelines.
- Custom domains — connect your own domain at justpublish.ai.
About this repository
Public metadata + registry listing for the Just Publish MCP server: server.json,
.mcp.json, icons, and this README. The server runs at
https://mcp.justpublish.ai/; its source is not part of this repo.
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。