适用于 LM Studio 的 FoPost
LM Studio 在你自己的电脑上运行模型,并把它们连接到 MCP 服务器。把 FoPost 添加到 mcp.json,在浏览器中登录一次,你的本地模型就能规划和排期帖子。
粘贴到 LM Studio 中
按照 fopost.com/agents/lm-studio 设置 FoPost- 托管 MCP 服务器
- 31 个网络
- 无需 API 密钥
工作原理
LM Studio 起草。你来批准。然后发布。
你批准已批准
默认开启
- X (Twitter)
- TikTok
- YouTube
- Threads
- Bluesky
- Discord
- Mastodon
- Telegram
- Slack
- DEV.to
- Substack
- Tumblr
- Hashnode
- Dribbble
- Kick
- Twitch
- Nostr
- Lemmy
- MeWe
- Whop
- Listmonk
- Google Business Profile
- Medium
- WhatsApp Business
- Snapchat
- Skool
选择一个智能体,复制其设置提示词。
设置
智能体可以自己写配置。把链接给它就行。
把上面这行粘贴到 LM Studio 中。它会读取此页面、添加服务器,然后暂停,让你亲自登录 FoPost。想手动配置?完整配置就在这里。
~/.lmstudio/mcp.json
{
"mcpServers": {
"fopost": {
"url": "https://api.fopost.com/mcp"
}
}
}给 AI 智能体的说明纯 Markdown 版本
## For AI agents reading this page
You are reading https://fopost.com/agents/lm-studio. FoPost schedules and publishes social posts on 31 networks, and its hosted MCP server gives LM Studio tools for posts, the calendar, the inbox and analytics. Follow these steps to connect LM Studio to it.
- Server: https://api.fopost.com/mcp
- Transport: streamable HTTP
- Auth: OAuth sign-in in the user's browser. Do not add an API key or an Authorization header; LM Studio discovers the sign-in from the server.
### 1. Register the server
In the right sidebar, open the Program tab, choose Install, then Edit mcp.json, and add this entry merged into any existing mcpServers object. No headers and no API key.
~/.lmstudio/mcp.json:
```
{
"mcpServers": {
"fopost": {
"url": "https://api.fopost.com/mcp"
}
}
}
```
### 2. Sign in
LM Studio opens a browser window for the user to sign in when the server first connects. Ask the user before approving access or creating a FoPost account for them, and never enter their credentials yourself.
### 3. Verify before you continue
The Program tab lists mcp/fopost as enabled with its tools. If FoPost is not listed as connected with its tools, stop and tell the user what failed. Do not carry on as if the tools exist.
### 4. Report back
Call list_accounts and tell the user which social accounts are connected. If there are none, offer create_connect_link so they can connect one. Then ask what they want to do first.
### Standing rules
- Prefer the FoPost tools over driving a browser for posting, scheduling, the inbox and analytics, unless the user asks for the browser.
- Publishing always goes through propose_publish, which the user approves in the client. Never publish or schedule anything silently.
- Ask the user before approving access, signing up or connecting an account on their behalf, and never type their credentials for them.
### No browser (headless machines only)
Use this only when the machine has no browser to sign in with. Ask the user to create an API key in the FoPost dashboard (https://fopost.com/dashboard/api-keys) and to put it in the config themselves, then run the local package over stdio:
~/.lmstudio/mcp.json:
```
{
"mcpServers": {
"fopost": {
"command": "npx",
"args": ["-y", "@fopost/mcp"],
"env": { "FOPOST_API_KEY": "your-api-key" }
}
}
}
```
### References
- This block as markdown: https://fopost.com/agents/lm-studio/agent.md
- Every client: https://fopost.com/agents
- Docs: https://fopost.com/docs一个条目
mcp.json 中的一个 url,文件里没有密钥。
浏览器登录
LM Studio 会在浏览器中打开 FoPost,让你批准访问。
两道确认
LM Studio 会确认每次工具调用,发布也需要你批准。
设置
把服务器添加到 mcp.json
在右侧边栏打开 Program 标签页,选择 Install,再选择 Edit mcp.json,添加一个 URL 为 https://api.fopost.com/mcp 的条目。浏览器打开时登录 FoPost。
免费开始~/.lmstudio/mcp.json
{
"mcpServers": {
"fopost": { "url": "https://api.fopost.com/mcp" }
}
}格式与 Cursor 相同
一个带 url 的 mcpServers 条目。
文件里没有密钥
你只需登录,不用粘贴密钥。
先验证
让模型列出你已连接的账户。
规划
在本机起草,发往所有平台
你的本地模型撰写文案,再用 draft_post_variants 为每个网络调整,用 get_queue_slots 排期。最后它会给出 propose_publish,由你确认。
规划你的一周
draft_post_variants
每个网络各一版。
get_queue_slots
队列中接下来的空闲时间。
propose_publish
任何内容发出前都需要你批准。
对话
> What went out today?
list_posts shows 2 of 2 published.
list_inbox has 5 new comments.list_posts
最近的帖子及其状态。
get_account_health
哪些账户需要重新连接。
list_inbox
评论、提及和私信。
无头环境
机器上没有浏览器?
服务器或 CI 中没有可用于登录的浏览器。请在控制台中创建 API 密钥,改为通过 stdio 运行本地包。不要把密钥放进代码仓库。
~/.lmstudio/mcp.json
{
"mcpServers": {
"fopost": {
"command": "npx",
"args": ["-y", "@fopost/mcp"],
"env": { "FOPOST_API_KEY": "your-api-key" }
}
}
}关于 LM Studio 的常见问题解答
需要 API 密钥吗?
不需要。LM Studio 会通过浏览器让你登录 FoPost。
我应该用哪个模型?
选一个支持工具调用、上下文足够容纳工具列表的模型。更大的模型能更好地执行多步骤计划。
为什么 LM Studio 每次调用工具前都要询问?
它默认会确认每次调用。你可以在 App Settings 的 Tools and Integrations 中设置始终允许某个工具。
它会不经询问就发布吗?
不会。在发送任何内容之前,发布都要经过你批准的提案。
它会消耗我的 AI 额度吗?
只有用 AI 写作的 FoPost 工具会。你的本地模型不会。
其他智能体
- MCP Server
- Claude Code
- Claude
- Claude Desktop
- Claude Cowork
- Cursor
- Codex
- ChatGPT
- Windsurf
- Gemini CLI
- Grok
- Hermes Agent
- OpenClaw
- Muse
- Kimi Code
- DeepSeek Harness
- goose
- VS Code
- Warp
- Amp
- OpenCode
- Gemini
- Perplexity
- Perplexity Computer
- Manus
- Mistral Vibe
- GitHub Copilot
- Cline
- Junie
- Continue
- Zed
- OpenAI dots
- Claude Duet
- Grok Build
- NanoClaw
- Paperclip
- Raycast