原型 · 预览版:收录官方注册中心里的全部服务器,每 30 分钟同步一次。评价和账户功能将在后续阶段推出。
PopNx MCP为使用 AI 的人准备的 MCP 服务器 · 用大白话讲清楚

Pixapi

作者:Pixapi-AI版本 0.1.4i★ –i电脑或云端
注册名称io.github.Pixapi-AI/pixapi-mcp

安装

概览

作者提供的说明(英文原文)

MCP server for Pixapi: check live credit pricing and balance, then generate images and video.

PopNx MCP 的三大支柱

这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。

它能帮你做什么

我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。

安装

选择你使用的 AI 应用,然后复制设置。

安装之前

你需要

  • 在云端运行无需安装任何东西,通过互联网连接
  • npm已安装 Node.js(用于 npx 命令)

你用的是哪个应用?

在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。

适用于网页版、桌面版和手机版 Claude:

  1. 打开 Settings → Connectors
  2. 点击 Add custom connector
  3. 粘贴下面的地址,然后点击 Add
  4. 如果要求登录,请登录作者的服务
SettingsConnectors1+ Add2https://…3
https://api.pixapi.ai/mcp

或者在你的电脑上运行

打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json

{
  "mcpServers": {
    "pixapi": {
      "command": "npx",
      "args": [
        "-y",
        "pixapi-mcp"
      ]
    }
  }
}

在终端中运行这条命令

claude mcp add --transport http pixapi https://api.pixapi.ai/mcp

或者在你的电脑上运行

claude mcp add pixapi -- npx -y pixapi-mcp

在 Cursor 中打开

或者自己复制设置

粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json

{
  "mcpServers": {
    "pixapi": {
      "url": "https://api.pixapi.ai/mcp"
    }
  }
}

或者在你的电脑上运行

在 Cursor 中打开

或者自己复制设置

粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json

{
  "mcpServers": {
    "pixapi": {
      "command": "npx",
      "args": [
        "-y",
        "pixapi-mcp"
      ]
    }
  }
}

浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。

在 VS Code 中打开

或者自己复制设置

粘贴到项目中的 .vscode/mcp.json

{
  "servers": {
    "pixapi": {
      "type": "http",
      "url": "https://api.pixapi.ai/mcp"
    }
  }
}

或者在你的电脑上运行

在 VS Code 中打开

或者自己复制设置

粘贴到项目中的 .vscode/mcp.json

{
  "servers": {
    "pixapi": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "pixapi-mcp"
      ]
    }
  }
}

浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。

ChatGPT 只能连接互联网上的服务器:

  1. 打开网页版 ChatGPT → Settings → Connectors (Apps)
  2. 点击 Create 或 Add(部分套餐需要先在 Advanced 中开启 Developer mode)
  3. 粘贴下面的地址,起个名字,然后保存
  4. 如果要求登录,请登录作者的服务
SettingsConnectors1+ Create2https://…3
https://api.pixapi.ai/mcp

菜单名称可能因套餐和版本不同而略有差异。

你的选择通过 Cookie 保存在这台设备上,无需账户。

这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。

它能做什么

我们还没有读取这个服务器的工具列表;我们会轮流检查各个端点。

基础检查i

我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月7日

  • ✓ 通过

    作者与仓库一致

    注册名称已通过 GitHub 账户 Pixapi-AI 验证,该账户拥有此仓库

  • … 尚未检查

    仍在维护

    我们还没有从 GitHub 读取这个仓库

  • … 尚未检查

    有许可证

    我们还没有从 GitHub 读取这个仓库

  • … 尚未检查

    端点有响应

    尚未检查;我们会轮流检查各个端点

  • … 尚未检查

    使用 OAuth 登录

    尚未检查;我们会轮流检查各个端点

  • … 尚未检查

    工具描述中没有隐藏指令

    尚未检查;我们会在检查端点时读取工具列表

与“设计、图片与音频”中的其他服务器比较i

  • 星标数多于此分类中其他 2186 个服务器的 0%
  • 发布到注册中心的时间比此分类中其他 2186 个服务器的 30% 更新
  • 基础检查结果等于或优于此分类中其他 2186 个服务器的 100%

与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。

常见问题

Pixapi 安全吗?

它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。

它需要哪些密钥或代码?

注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。

它在我的电脑上运行还是在云端?

两种方式都可以:安装在你的电脑上,或者连接云端版本。

它还在维护吗?

GitHub 仓库最后更新于–,最新版本(0.1.4)于2个月前进入注册中心。

哪些应用可以使用它?

支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。

有什么可以替代它?

同一分类中有:Figma Context MCP, Claude Real Video, Figma MCP Server

替代选择

同一分类中其他作者的服务器,按星标数排序。

如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i

Figma Context MCP

作者:GLips

Give your coding agent access to your Figma data. Implement designs in any framework in one-shot.

在你的电脑上运行★ 16,165MIT5个月前更新
安装

Claude Real Video

作者:HUANGCHIHHUNGLeo

Let any LLM watch a video locally — and search everything it has ever watched.

在你的电脑上运行★ 2,284MIT2个月前更新
安装

Figma MCP Server

作者:figma

The Figma MCP server brings Figma design context directly into your AI workflow.

在云端运行★ 2,15213个月前更新
安装

作者的 README

打开 README

按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。

pixapi-mcp

Use Pixapi image and video generation tools from MCP clients.

  • Website: Pixapi
  • Remote MCP endpoint: https://api.pixapi.ai/mcp
  • License: MIT

Quick start

Configure your project automatically (Node.js 22+):

npx -y pixapi-mcp init --client codex

Replace codex with your client, or use --client all for all five:

Client--clientConfiguration file
Claude Codeclaude-code.mcp.json
Cursorcursor.cursor/mcp.json
Codexcodex.codex/config.toml
VS Code / GitHub Copilotvscode.vscode/mcp.json
Gemini CLIgemini-cli.gemini/settings.json

Without --client, the default is Claude Code and Cursor (--client both). Use --project /path/to/project to configure another project.

Reload your MCP client and connect to pixapi. Complete its browser sign-in and authorization prompt the first time. The client saves the OAuth session and refreshes it automatically; no API key or credential editing is required. The client may ask you to sign in again if authorization is revoked or expires. Each client manages its own native OAuth session.

init configures a direct remote HTTP connection by default. If your client has an Add MCP Server interface, you can also enter the endpoint URL there without installing this npm package or Node.js.

Codex requires a trusted project. In VS Code, use MCP: List Servers to start pixapi; in Gemini CLI, use /mcp to inspect it. Complete any client trust or sign-in prompt. init does not change client security policies.

Generated configuration

Claude Code:

{"mcpServers":{"pixapi":{"type":"http","url":"https://api.pixapi.ai/mcp"}}}

Cursor uses the same mcpServers entry with url and no type field. VS Code uses servers with type: "http" and url. Gemini CLI uses mcpServers with httpUrl.

Codex:

[mcp_servers.pixapi]

url = "https://api.pixapi.ai/mcp"

Running init updates the pixapi entry and preserves other servers and settings. VS Code JSONC comments are preserved. Codex TOML settings are preserved, but comments and formatting are not. All selected configurations are parsed before writing; a filesystem failure during writes can still leave some files updated. Correct the local issue and rerun init.

Local stdio bridge

For clients that need stdio, configure the bridge explicitly:

npx -y pixapi-mcp init --client cursor --transport stdio

The generated entry runs npx -y pixapi-mcp proxy without credentials. Registry installations of the npm package also start the bridge directly: no separate init command or environment variable is required.

The bridge completes the local MCP handshake immediately. On the first tool request it opens browser authorization, then forwards tools to Pixapi. Subsequent connections reuse the saved session and refresh tokens automatically. For clients with a short tool-discovery timeout, sign in once beforehand:

npx -y pixapi-mcp login

If a client times out during the initial browser sign-in, complete sign-in and reconnect the client. No authorization URL, code, or token is printed to MCP stdout or error output.

OAuth sessions are stored per endpoint in ~/.pixapi/oauth- .json. On macOS and Linux, the directory uses mode 0700 and files use 0600. Writes are atomic; concurrent bridge processes share an authorization lock. PKCE verifiers and callback state are kept only in memory. The callback binds to loopback on the same host as the bridge. For a headless or remote host, prefer the client's native remote OAuth connection.

Existing API key installations

Existing API key configurations remain compatible. For an explicit API key installation, set PIXAPI_API_KEY when running init; it configures stdio and saves the key privately in ~/.pixapi/mcp-credentials.json. An explicit --transport remote always selects native OAuth, even if that variable is set. The proxy also accepts PIXAPI_API_KEY directly for unattended deployments.

The proxy gives an explicit environment key priority, then a saved OAuth session, then a legacy key file. To migrate an existing bridge to OAuth, run pixapi-mcp login and remove any PIXAPI_API_KEY override from its environment. Keys and tokens never appear in generated project configurations.

Available tools

This package forwards the live tool list from the remote Pixapi MCP endpoint. After you connect, call tools/list or pixapi_get_pricing_and_balance for the current catalog.

pixapi_get_pricing_and_balance

Returns your current balance and live image/video model pricing. Each catalog row includes a type:

  • Images: text_to_image, image_to_image
  • Videos: text_to_video, image_to_video

Rows marked mcp_generate_supported=true can be called through the matching generate tool.

pixapi_generate_image

Starts an asynchronous image task and returns a task_id. Use type=text_to_image for a prompt-only request, or type=image_to_image with the image field. The tool accepts the image models in the live catalog, including Gemini, GPT Image, Flux, Seedream, Qwen, and Wan.

This call consumes account credits. It is not idempotent: repeating the same call creates another task and may charge the account again.

pixapi_generate_video

Starts an asynchronous video task and returns a task_id. Use type=text_to_video for a prompt-only request, or type=image_to_video with the image field. Pass seconds and resolution as required by the selected model.

This call consumes account credits and is not idempotent.

pixapi_get_task

Retrieves an image or video task owned by the authenticated account. Poll this tool with the returned task_id until the status is completed or failed. A completed task includes media URLs.

Recommended agent flow:

  1. Call pixapi_get_pricing_and_balance.
  2. Select a catalog row with mcp_generate_supported=true and check the balance.
  3. Call pixapi_generate_image or pixapi_generate_video once and retain its task_id.
  4. Poll pixapi_get_task every few seconds.
  5. Return the result URL when the task is complete.

CLI reference

pixapi-mcp init [--client <name|all|both>] [--project <path>] [--transport <remote|stdio>]

pixapi-mcp login pixapi-mcp proxy pixapi-mcp --help

With no arguments, the package starts the stdio proxy.

Troubleshooting

  • Tools do not appear: reload the client, enable the pixapi server, and complete project trust and browser sign-in. For stdio, ensure Node.js 22+ and npx are available on the host running the bridge.
  • Authorization denied or timed out: retry sign-in from the native client, or run pixapi-mcp login for the bridge and complete browser authorization.
  • Callback port unavailable: close another pending Pixapi login and retry.
  • Service unavailable: check your connection and retry later. Internal HTTP responses and exception details are omitted from public errors.
  • Invalid legacy API key: replace the key, or migrate to OAuth as above.

Security and current limitations

  • Do not commit or share files in ~/.pixapi/.
  • Revoke unused or exposed credentials in your Pixapi account.
  • The stdio bridge forwards tools only. Resources, prompts, sampling, and elicitation are not exposed by this package.
  • Image and video generation are asynchronous and not idempotent. The bridge retries HTTP authentication rejection once; it does not retry network failures or server errors that might have occurred after a tool executed.
  • Account top-up is not exposed as an MCP tool; complete payment on Pixapi.

For service documentation, visit Pixapi docs.

Official MCP Registry

This server is published to the official MCP Registry as io.github.Pixapi-AI/pixapi-mcp, with both connection methods declared in server.json: the remote endpoint and the npm package. Both support OAuth without a required API key starting with version 0.1.4.

curl "https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.Pixapi-AI/pixapi-mcp"

Set matching versions in package.json, package-lock.json, and server.json before pushing the corresponding v* tag. The release workflow installs locked dependencies, runs checks, and validates the Registry manifest before publishing. Manual runs must also select that release tag. A rerun compares existing npm package integrity and Registry metadata and only publishes missing steps; conflicting contents require a new version. Registry authentication uses GitHub OIDC, without a stored Registry token.

问 PopNx