micropub-mcp

micropub-mcp

Enables AI agents to create, update, delete, and query posts on any Micropub-compatible blog with IndieAuth authentication.

Category
访问服务器

README

micropub-mcp

MCP server that lets AI agents (Claude Code, Cursor, etc.) create, update, delete, and query posts on any Micropub-compatible blog. Authenticates via IndieAuth with PKCE.

Built with Bun + TypeScript.

Quick Start

1. Install

git clone https://github.com/rmdes/micropub-mcp.git
cd micropub-mcp
bun install

2. Add to Claude Code

Add to your .mcp.json (project or global):

{
  "mcpServers": {
    "micropub": {
      "command": "bun",
      "args": ["run", "/path/to/micropub-mcp/src/index.ts"]
    }
  }
}

3. Authenticate

Tell Claude: "Authenticate with my blog at https://yourblog.com"

This calls micropub_auth, which:

  1. Discovers your blog's IndieAuth endpoints
  2. Opens your browser to the authorization page
  3. You approve access, the callback server catches the token
  4. Token is saved to ~/.config/micropub-mcp/<domain>.json

Authentication persists across sessions. You only need to do this once.

4. Start Posting

"Create a note saying Hello from my MCP client!"
"Write an article titled 'My Setup' about my development environment"
"Syndicate a note to Bluesky and Mastodon about the IndieWeb"

Tools

micropub_auth

Authenticate with a Micropub-compatible blog via IndieAuth + PKCE.

Parameter Type Required Description
site_url string Yes Your blog URL (e.g., https://yourblog.com)
scope string No OAuth scopes (default: create update delete media)

micropub_create

Create a new post (note, article, photo, bookmark, etc.).

Parameter Type Required Description
content string No Post content (text or HTML)
type string No Post type: entry (default), event
name string No Title (makes it an article)
summary string No Post summary
category string[] No Tags/categories
syndicate_to string[] No Syndication target UIDs
in_reply_to string No URL being replied to
like_of string No URL being liked
repost_of string No URL being reposted
photo string[] No Photo URLs
slug string No URL slug
post_status string No published or draft
published string No ISO 8601 date
ai_text_level string No AI text involvement: 0 (None), 1 (Editorial), 2 (Co-drafting), 3 (AI-generated)
ai_code_level string No AI code involvement: 0 (Human), 1 (AI-assisted), 2 (Primarily AI)
ai_tools string No AI tools used (e.g. "Claude", "ChatGPT, Copilot")
ai_description string No Free-text description of AI usage

Examples:

Create a note: { content: "Hello world" }
Create an article: { name: "My Title", content: "Article body..." }
Create a bookmark: { like_of: "https://example.com/post" }
Create a reply: { in_reply_to: "https://example.com/post", content: "Great post!" }
Syndicate: { content: "Cross-posted!", syndicate_to: ["https://brid.gy/publish/bluesky"] }
With AI metadata: { content: "AI-assisted post", ai_text_level: "2", ai_tools: "Claude" }

micropub_update

Update an existing post.

Parameter Type Required Description
url string Yes URL of the post to update
replace object No Properties to replace (values are arrays)
add object No Properties to add
delete_properties string[] No Property names to remove

Example: { url: "https://blog.com/notes/abc", replace: { content: ["Updated text"] }, add: { category: ["new-tag"] } }

micropub_delete

Delete a post.

Parameter Type Required Description
url string Yes URL of the post to delete

micropub_undelete

Restore a previously deleted post.

Parameter Type Required Description
url string Yes URL of the post to restore

micropub_query

Query the Micropub server for configuration, posts, or metadata.

Parameter Type Required Description
q string Yes Query type (see below)
url string No Post URL (for source queries)
properties string[] No Properties to return
limit number No Max results
offset number No Pagination offset

Query types:

  • config — Server capabilities, syndication targets, post types
  • source — Get a post's properties (requires url)
  • syndicate-to — Available syndication targets
  • post-types — Supported post types
  • category — Available categories/tags
  • channel — Available channels

micropub_upload

Upload a media file to the server's media endpoint.

Parameter Type Required Description
file_path string Yes Absolute path to the file

Returns the uploaded file's URL, which can be used in photo, video, or audio parameters of micropub_create.

AI Transparency Metadata

Posts can optionally carry metadata disclosing AI involvement. These fields are sent as standard Micropub mf2 properties (ai-text-level, ai-code-level, ai-tools, ai-description).

All fields are optional. If your Micropub server doesn't know about them, it will simply store them as extra properties or ignore them — no special server-side support is required. This works with any standard Micropub endpoint.

Level ai_text_level ai_code_level
0 None — human wrote everything Human-written
1 Editorial assistance (grammar, spelling) AI-assisted
2 Co-drafting (human directed, AI wrote) Primarily AI-generated
3 AI-generated (human reviewed)

If your server does support these fields (e.g., Indiekit with @rmdes/indiekit-endpoint-posts), they can be displayed in the post form UI and used for filtering or labeling posts.

How Authentication Works

┌─────────┐     ┌──────────────┐     ┌─────────────┐     ┌──────────┐
│  Claude  │────>│  MCP Server  │────>│   Browser    │────>│   Blog   │
│  Code    │     │  (this tool) │     │              │     │  Server  │
└─────────┘     └──────────────┘     └─────────────┘     └──────────┘
     │                │                     │                    │
     │  micropub_auth │                     │                    │
     │───────────────>│                     │                    │
     │                │  discover endpoints │                    │
     │                │────────────────────────────────────────>│
     │                │  start callback     │                    │
     │                │  server :19750      │                    │
     │                │  open browser ──────>│                    │
     │                │                     │  authorize ────────>│
     │  "open browser"│                     │                    │
     │<───────────────│                     │<── redirect with   │
     │                │                     │    auth code       │
     │                │<── callback ────────│                    │
     │                │  exchange code ─────────────────────────>│
     │                │<── access token ────────────────────────│
     │                │  save token         │                    │
     │                │  ~/.config/...      │                    │

The callback server runs on localhost:19750 and is a process-level singleton — it persists across MCP tool calls and shuts down after receiving the callback.

Token Storage

Tokens are saved to ~/.config/micropub-mcp/<domain>.json and include:

  • Access token
  • Refresh token (if provided)
  • Expiration time
  • Discovered endpoints (micropub, media, token)

Expired tokens are automatically refreshed if a refresh token is available.

Requirements

Indiekit-Specific Notes

If your blog runs Indiekit behind nginx, ensure:

  1. CSP form-action: The /auth location must allow form-action * (or at minimum http://localhost:*) so the OAuth consent form can redirect to the local callback server. Without this, clicking "Allow" does nothing (HTTP 499 in logs).

  2. Redirect validation: Indiekit's upstream redirect regex (/^\/[\w&/=?]*$/) rejects hyphens in paths like /auth/new-password. If you hit ForbiddenError: Invalid redirect attempted, patch lib/indieauth.js with an expanded regex.

Development

# Run tests
bun test

# Start the MCP server (stdio)
bun run src/index.ts

Architecture

src/
├── index.ts        Entry point — creates MCP server, connects stdio transport
├── tools.ts        7 MCP tool definitions with Zod schemas
├── auth.ts         IndieAuth + PKCE flow, token storage, non-blocking callback server
├── client.ts       Micropub HTTP client (create/update/delete/query/upload)
└── discovery.ts    Endpoint discovery from Link headers + HTML + indieauth-metadata

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