
HumanPost
io.github.HumanPost-co由 GitHub 账号 HumanPost-co 发布v1.0.2更新于 Nov 6, 2026
Create social drafts, queue approved posts, and track human posting and analytics in HumanPost.
发布方自己的服务:TAP
概览
托管式 HumanPost MCP 服务器,让助手创建草稿、排队已批准的帖子,并跟踪由 HumanPost 人工运营完成的 TikTok 和 Instagram 发布与数据。
- 可以做什么
- 查看连接上下文、所属公司与权限范围,并列出可用的社交账号
- 查看授权公司的发布队列、草稿、任务与分析数据
- 上传媒体并准备视频或轮播草稿供审核,必要时请求预览
- 在授权范围内排队内容,并跟踪帖子或任务状态直至获得发布链接
- 适用场景
- 适合希望通过托管发布服务准备和安排 TikTok 或 Instagram 内容,或在对话中查看账号开通、队列与表现的场景。最适合已经在使用 HumanPost 人工发布服务的团队。
- 运行要求
- 远程 Streamable HTTP 端点 humanpost.com/mcp,无需本地安装包。需要 HumanPost 账号及公司访问权限。认证方式为 OAuth 2.1(PKCE、动态客户端注册),或使用公司级 API 密钥作为 Bearer 令牌,例如通过 HUMANPOST_API_KEY 环境变量提供。密钥在控制台创建,初始权限为 accounts:read。
安装
在 SourceWeft 中
- 打开 控制台中的 HumanPost,将其添加到工作区。
- 连接 所需的 凭据。它们会被私密存储,绝不会在公开页面上展示。
- 为需要使用其工具的对话启用该服务。
Web executable,通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。
交给智能体安装
把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体,或 SourceWeft 对话中。智能体会读取该服务器的安装指南,先向你展示发布者、端点和登录方式,经你同意后再把它添加到你的客户端。密钥不会写进任何命令。
阅读 https://sourceweft.com/mcp/io.github.HumanPost-co%2Fhumanpost/install.md,并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前,先告诉我它的发布者、端点和登录方式,等我确认。不要把 API 密钥写进命令或我会提交的文件;问我密钥应从哪里获取。其他 MCP 客户端
把它添加到你客户端的 mcpServers 配置中。 客户端会在首次连接时完成 OAuth 登录。
{
"mcpServers": {
"humanpost": {
"type": "http",
"url": "https://humanpost.com/mcp"
}
}
}README
HumanPost MCP and agent plugins
Official integration package for HumanPost: request TikTok and Instagram account setup, prepare video and carousel drafts, and coordinate posting at scale through HumanPost’s human operators. Track account setup, posting progress, published links and performance from your AI assistant.
HumanPost operates a hosted MCP service. This repository contains client configuration and a publishing skill; it does not contain the hosted server implementation. A HumanPost account and company access are required. OAuth-capable clients use HumanPost sign-in; company-scoped API keys remain available for manual configurations. Service usage and some actions may incur charges under your HumanPost plan.
Connect
- Transport: Streamable HTTP
- Endpoint:
https://humanpost.com/mcp - Authentication: OAuth 2.1 authorization code with S256 PKCE and dynamic client registration. Company-scoped API keys are also supported.
- Setup and tools · API reference · Complete agent documentation
Create a key under Dashboard → Settings → API & MCP. Start with accounts:read and add only the scopes your workflow needs. Each key belongs to one company. Store it in your client's secret configuration; never commit it or paste it into a conversation. OAuth users select one company and minimum permissions in the consent screen. Writes are off by default. Access tokens expire after 15 minutes; refresh tokens rotate and expire after seven days. Revoke a connection under Settings → API Keys. No credential is included in this package.
Cursor
The Cursor plugin uses the canonical OAuth endpoint. Install it and use Cursor’s Connect action to sign in to HumanPost. For manual setup:
If you use a company API key instead, configure an Authorization header in your private Cursor settings with Bearer ${env:HUMANPOST_API_KEY}.
Cline
Merge config-cline.json into Cline's MCP settings. Replace YOUR_HUMANPOST_API_KEY in your local settings with your scoped key. Keep that settings file private. The transport value is streamableHttp; keep autoApprove empty and review tool calls. Cline CLI uses ~/.cline/data/settings/cline_mcp_settings.json by default and supports CLINE_MCP_SETTINGS_PATH for an isolated configuration.
Claude Code
Add this repository as a marketplace, then install the plugin:
Set HUMANPOST_API_KEY in the process environment before launching Claude Code. Alternatively, use the .mcp.json entry in your project. The compatibility .mcp.json uses an environment-supplied API key. For native OAuth instead, run claude mcp add --transport http humanpost https://humanpost.com/mcp and use Claude Code’s MCP authentication flow. Claude web directory approval is separate from installing this plugin.
Grok Build
Grok Build supports Claude-compatible plugins. Add the repository using its plugin marketplace command:
Install humanpost from the Marketplace tab and supply HUMANPOST_API_KEY in the client environment. The package includes both .grok-plugin/plugin.json and .claude-plugin/plugin.json. Review the configured MCP before enabling it. Official marketplace inclusion is separate from adding this publisher's marketplace.
Grok Bot
Grok Bot has a separate app-directory submission form, which also covers Cursor. The Grok Build CLI instructions above apply to Build; they do not install the Bot app plugin. This repository includes the Cursor marketplace manifest and the canonical OAuth MCP entry for app-directory review.
Ask your Bot to add a custom MCP server called HumanPost at https://humanpost.com/mcp, then use its Authorize card. In HumanPost, choose one company and the minimum scopes needed; writes are off by default. Start with accounts:read. Grok Bot connections are account-level, so revoke a test connection in HumanPost Settings → API Keys and remove the temporary connector when finished.
Marketplace approval is separate from adding a private custom connection. The Grok Bot-specific application was submitted and acknowledged on October 24, 2026; marketplace review is pending. No approved Bot listing is claimed.
Muse Code
Merge the mcp_servers.humanpost entry from config-muse.json into your Muse Code settings. It uses Muse's documented streamable_http transport and ${HUMANPOST_API_KEY} environment interpolation. Install the publishing skill with:
Muse Code configuration is separate from listing a connector in the consumer Muse directory. No consumer directory approval is claimed.
Agent Skills / skills.sh / OpenClaw
The skill uses a connected MCP or the documented REST API on hosts that can make HTTP requests. It cannot supply network or shell capabilities that the host does not have. Installing the skill does not configure credentials or grant company access.
ChatGPT and Codex
The root plugin.json and mcp.json use the portable Agent Plugins format. The portable MCP entry uses the OAuth endpoint and discovers its authorization server.
The latest complete public upload bundle is humanpost-openai-plugin-1.0.3.zip from the 1.0.3 release. The 1.0.4 review bundle was provided directly to OpenAI; the 1.0.4 release publishes only the approved walkthrough. The upload bundle's root mcp.json uses the documented OpenAI authentication extension from openai/mcp.json: OAuth with DCR and read-only default scopes. This separate bundle preserves strict portable-schema compatibility for other hosts. No client secret, access token or API key is bundled.
The endpoint is deployed and its browser consent, scope enforcement, refresh and revocation were live-tested using an isolated test company. OpenAI business verification and publisher setup are complete. HumanPost metadata and skills version 1.0.4 was submitted on October 24, 2026, and the portal confirms In review. It is not yet an approved or published listing.
To reproduce a published release, check out its matching Git tag first, then build both archives with python3 scripts/build_release.py --output-dir /tmp/humanpost-release. Later documentation commits do not alter previously published archives.
Typical workflow
- Call
get_contextto check company and scopes, thenlist_accountsto select the correct account. - Inspect accounts, posting queues, drafts, or analytics using the discovered tools.
- Upload media and prepare a draft for review. Request an actual preview when needed.
- Queue content only within the user's authorization. Preserve idempotency keys and returned operation IDs on retries.
- Inspect the resulting post/task status. Queued and pending-approval states are not confirmed publication.
The server advertises its tool catalog and enforces key scopes when each tool is called. Discovery alone does not grant permission to invoke a tool. Account provisioning, task creation, rendering, posting, schedule changes, and cancellations may cause external effects or consume credits. Do not auto-approve every tool. Availability and schemas come from live tool discovery; see the public documentation for details and limitations.
Security and network access
OAuth clients connect to humanpost.com/mcp and discover authorization endpoints on the same HumanPost origin. The compatibility API-key configs send your key only to us-central1-clickbaitcarousel.cloudfunctions.net/mcp, HumanPost's Firebase-hosted service. The REST fallback uses the same host under /publishingApi. Documentation links go to humanpost.com. Uploads may use service-issued signed storage URLs; never send the API key to those URLs. The package has no lifecycle hooks, background processes, telemetry, dependency installers, or remote-code execution.
For a read-only health check without a key, python3 scripts/check_endpoint.py verifies that the public endpoint rejects unauthenticated initialization. An authenticated connection test is available in the HumanPost dashboard. A successful unauthenticated check does not validate every tool or company workflow.
Privacy and support
The hosted service processes account, content, task, and analytics data for the company authorized by your connection or API key. See the HumanPost Privacy Policy and Terms of Service. Contact [email protected] or open an integration issue. Never include credentials or private customer content in a public issue.
The integration package is MIT licensed. The hosted HumanPost service and HumanPost branding are not licensed by that grant. Platform compatibility configurations and a submission do not imply marketplace endorsement or approval.
OpenAI review materials
The 1.0.4 manifest includes five positive cases, three negative cases, release notes, and a link to an edited screen recording of the actual ChatGPT tests. The video contains sample-company data only; credential entry and unrelated account information are excluded. Idle time is shortened. Read/refusal cases were run with a read-only connection; draft creation was run after separate write consent. Both test grants were revoked and disconnected.
These cases exercise the hosted MCP through private ChatGPT test connectors. Publisher package and skill checks are separate; they do not imply that every advertised workflow (account setup, queueing, live social publication, or tasks) was executed. Reviewer login details are supplied only through OpenAI's private review form and are never bundled here.
来源:README.md,提交 0377577
工具
0版本历史
2- v1.0.2最新Oct 23, 2026
- v1.0.1Oct 23, 2026
评分
已满足适用检查项的 85%
该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。
- 近 30 天内可连通必需25 分
- 已声明认证方式必需10 分
- 已列出工具15 分
请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。
- 描述必需15 分
- 图标10 分
- 文档15 分
- 许可证10 分
其他 Human 集成

TAP
tech.human官方发布者
TAP 是一个托管式远程 MCP 端点,通过占位密钥、策略校验和可选的人工审批,为 AI 代理隔离凭据。
UGC, the Human Ad Network
africa.ugc官方发布者
UGC(“Human Ad Network”)的远程 MCP 端点,让助手在对话中处理广告活动、剪辑与 Clip Studio。

Woodstitch
com.woodstitch
对文本转语音音频发起付费的盲听人工评测,返回带置信区间的偏好或人机判别结果。

J-Dawgs AI Tool Store & Human Hands — Catalog MCP
由 GitHub 账号 joshuajdawg0350 发布
J-Dawgs AI 工具商店与 Human Hands for AIs 的只读目录服务器,让助手搜索工具、价格和套餐,并了解智能体如何发布现实任务。

ActualHumanTake
由 GitHub 账号 magiautonomous 发布
提交产品以获得真人撰写的诚实评价,并让助手凭秘密回执令牌查询进度或读取已完成的反馈。