Instant Expert

io.github.Instant-Expertv3.0.0更新于 Oct 2, 2026

Find specific execs, operators and experts and invite them to paid calls. Pay only when they book.

已验证Streamable HTTP可网页运行Business & Commerce Productivity & Workflow Communication & Collaboration

概览

AI 生成的概览

查找特定专业人士并向其发送付费通话或书面回答邀请，仅在对方预约或回复时收费。

功能: Instant Expert 让助手研究并保存特定人群名单，例如高管或专家，并查找其工作邮箱。助手随后起草邀请，提供 15 至 60 分钟的付费通话，或对一个问题给出书面或语音回答，并设定每人价格和总支出上限。发送前你会看到订单预览并需批准。回复、预约时间和记录可在询问时返回给助手。
适用场景: 适合早期客户调研、针对某类人群的外联、用户测试与访谈，或处理你已有的潜在客户名单。当你需要特定专业人士的答复或对话，而非一般网络搜索时适用。
运行要求: 远程 Streamable HTTP 端点，使用 Instant Expert 账户进行 OAuth 登录；本地无需运行任何程序，也无需管理 API 密钥。作用域为 email 和 profile，以及用于刷新令牌的 offline_access；openid 和 phone 会被拒绝。付费发送需要已保存的支付卡。

安装前请注意

付费发送默认关闭，需按助手单独允许；即便如此，每笔订单仍需明确批准。支出为真实费用，每人最低 5 美元，并可通过 max_spend_cents 设置总上限。卡信息不会经过聊天，只能使用已保存在 instant.expert 的卡。搜索请求、联系人、消息和预算会发送给 Instant Expert，邮箱地址不会返回给助手。

安装

在 SourceWeft 中

打开控制台中的 Instant Expert，将其添加到工作区。
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。

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "instant-expert": {
      "type": "http",
      "url": "https://instant.expert/mcp"
    }
  }
}

README

[Instant Expert]

Instant Expert for AI assistants

Instant Expert gets you on the phone with specific people: the VP of Sales at a Series A startup, a head of procurement at a hospital system, a claims manager at a regional insurer. You describe who you want (or paste a lead list), it finds them and their work email, and it sends each one a paid invitation to a 15 to 60 minute call or a written or voice answer to one question. You pay only when someone books or replies.

This repo has the install files for using Instant Expert from Claude, Cowork, Claude Code, Cursor, VS Code, Codex, Gemini CLI and other MCP clients. The server itself is hosted at https://instant.expert/mcp, so there's nothing to run locally and no API key to manage; you sign in with your Instant Expert account the first time a tool runs.

MCP docs · Limits and costs · Support

What people use it for

Early customer discovery. "We're building expense software for construction firms. Who should we talk to?" The assistant turns that into a few concrete audiences (say, controllers at US general contractors with 50 to 500 employees), finds them and drafts the ask.
Outreach to a persona. "Find 30 heads of RevOps at Series B fintechs and invite them to a 15-minute call about pipeline tooling."
A lead list you already have. Paste LinkedIn URLs or emails (up to 100 per batch). Instant Expert finds work emails where needed and drafts the invitations.
User testing and interviews. Short paid calls or written answers from a specific kind of professional, for example "five security leads at companies that use Okta".
Following up. "Did anyone book yet? What did the people who replied say?"

How it works

You describe the people (or the problem you're trying to learn about) in plain language.
Instant Expert researches and saves a list. A research search often takes a few minutes, and the assistant tells you how many people it found against how many you asked for.
The assistant drafts the invitation: your message, call or written answer, price per person and a total cap.
You review an order preview (recipients, message, prices, cap, card and terms) and approve it. Nothing is sent or charged before that.
Invitations go out by email from Instant Expert. Replies, booked times and transcripts come back to the assistant when you ask.

Pricing

Searching, importing contacts and preparing drafts are free. They count toward your account's limits.
You set the price per person ($5 minimum, Instant Expert's fee included), or use each person's suggested price. Typical offers are probably in the $25 to $300 range, depending on seniority and whether it's a call or a written answer.
A call is charged when the person books it. A written or voice answer is charged when the reply is completed. If nobody accepts, you pay nothing.
A total cap (max_spend_cents) limits spend, so you can invite more people than the cap would cover if everyone said yes.

Install

The MCP endpoint for every client is:

It uses Streamable HTTP with OAuth sign-in. Paid sending is off for a new connection until you allow it (see Permissions).

Claude (claude.ai, desktop and mobile)

Plugin (connector plus the workflow skill), recommended:

Open Customize → Plugins → Add → Add marketplace and enter Instant-Expert/instant-expert-plugins.
Install Instant Expert.
Open the plugin's Connectors tab and connect Instant Expert. Sign in and approve.

Connector only: add Instant Expert as a custom connector, or go to Customize → Connectors → Add custom connector and paste the URL above. On Team and Enterprise plans an Owner adds connectors for the organization first.

Cowork

Same steps as Claude above: Customize → Plugins → Add → Add marketplace, enter Instant-Expert/instant-expert-plugins, install, then connect from the plugin's Connectors tab. A plugin you install is saved to your account, so it also shows up in claude.ai chat and in Claude Code.

Claude Code

text

/plugin marketplace add Instant-Expert/instant-expert-plugins/plugin install instant-expert@instant-expert

/plugin marketplace add Instant-Expert/instant-expert-plugins/plugin install instant-expert@instant-expert

Then run /mcp, pick instant-expert and choose Authenticate. Connector only, without the skill:

Cursor

Add to Cursor (or open cursor://anysphere.cursor-deeplink/mcp/install?name=instant-expert&config=eyJ1cmwiOiJodHRwczovL2luc3RhbnQuZXhwZXJ0L21jcCJ9), or add this to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):

Cursor shows the server as needing login; select it to sign in.

VS Code (GitHub Copilot agent mode)

Install in VS Code (or open vscode:mcp/install?%7B%22name%22%3A%22instant-expert%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Finstant.expert%2Fmcp%22%7D), or add this to .vscode/mcp.json:

json

{  "servers": {    "instant-expert": {      "type": "http",      "url": "https://instant.expert/mcp"    }  }}

{  "servers": {    "instant-expert": {      "type": "http",      "url": "https://instant.expert/mcp"    }  }}

From a terminal:

VS Code asks you to sign in the first time it starts the server.

Codex CLI

Add this to ~/.codex/config.toml. The scopes line matters: Instant Expert refuses the openid scope, which Codex otherwise requests.

Then sign in:

Gemini CLI

The extension adds the server and a GEMINI.md context file with the workflow. Restart Gemini CLI, then run /mcp auth instant-expert to sign in. (Since June 18, 2026, Gemini CLI serves Gemini Code Assist Standard and Enterprise and paid API keys; other users moved to Antigravity CLI.)

Grok

Go to grok.com/connectors, click New Connector, choose Custom, paste the URL above and sign in. On Grok Business and Enterprise, an admin provisions the connector first.

Perplexity

Open Account settings → Connectors, click + Custom connector, choose Remote, name it Instant Expert, paste the URL, pick OAuth and Streamable HTTP, then click the connector card to sign in. Enterprise admins can share it with the whole organization.

Le Chat (Mistral)

Open the side panel, expand Intelligence → Connectors, click + Add Connector, switch to the Custom MCP Connector tab, name it instant-expert, paste the URL and click Connect. Le Chat detects the OAuth sign-in. On Free, Pro and Student plans the account owner is the admin who can add connectors.

ChatGPT

ChatGPT uses a separate Instant Expert plugin from its own directory. It finds people and prepares drafts; you review and send them on instant.expert. See the ChatGPT docs.

Any other MCP client

Transport: Streamable HTTP at https://instant.expert/mcp.
Auth: OAuth 2.1 authorization code flow with PKCE and dynamic client registration. An unauthenticated request gets a 401 pointing to https://instant.expert/.well-known/oauth-protected-resource/mcp.
Scopes: email and profile, plus offline_access for refresh tokens. Requests for openid or phone are refused at consent.

If sign-in fails in a client we haven't listed, email [email protected] with the client name and the error.

Permissions

Every connection can search, import contacts, prepare drafts and read your searches, drafts, sent requests and replies.
Paid sending is off by default. You allow it per assistant on the consent screen or later in Connected assistants. Without it, the assistant prepares drafts and you send them from Requests.
Even with sending allowed, the assistant has to show you each order and get your explicit approval first. It can only use a card you've already saved on instant.expert; if you don't have one, it gives you a link to add one there. Card details never pass through chat.
You can disconnect an assistant at any time in Connected assistants. Each connection is its own OAuth grant.

Data and privacy

The files in this repo are Markdown and JSON; nothing here runs on your machine. When you use the tools, your assistant sends Instant Expert what you ask it to work with: search requests, contacts you provide, messages, budgets and order confirmations. Instant Expert uses them to run searches, find work emails at delivery time and deliver your invitations. Email addresses aren't returned to the assistant. See the privacy policy and terms.

What's in this repo