TGGET Niche Research

io.tggetv1.0.1更新于 Sep 29, 2026

Validate app and SaaS ideas: search demand, competitors, app store data and a verdict.

已验证Streamable HTTP可网页运行Business & CommerceData & AnalyticsWeb Search & Scraping

概览

AI 生成的概览

用实测搜索需求、竞品与应用商店数据以及规则化结论来验证 App 或 SaaS 创意。

功能
这是 TGGET 的远程服务器，可根据搜索词或用户自己描述的想法研究细分市场。工具支持启动并轮询研究、获取最终报告（结论、客户痛点、MVP、粗略经济测算、SEO 与 ASO 方案、风险）、追加单项信号（竞品、需求历史、App Store 评论、社区信号）、对比两到五个已完成的研究，以及关注某个细分市场并每月复查。还提供配额与数据来源资源，以及 find_niche 提示。
适用场景
适合在决定是否开发某个 App 或在线服务时，用需求与竞品证据代替猜测。也适合比较多个候选创意，或按月跟踪某个细分市场的变化。
运行要求
远程 Streamable HTTP 端点 tgget.io/mcp，无需本地运行时；仅支持本地服务器的客户端可用 npx mcp-remote 桥接。需要 TGGET 账号（免费账号即可）和个人访问令牌，通过 Authorization Bearer 请求头发送，或在 Codex 中使用 TGGET_TOKEN 环境变量。需要访问 tgget.io 的网络连接。
安装前请注意
需要 Authorization 请求头中的个人访问令牌；令牌只显示一次，一年后过期，修改密码或关闭会话即被撤销。免费方案的研究是公开的，可能被列入公开细分市场目录（仅显示研究词与结果，不显示你输入的描述）；付费方案保持私密。研究在启动时即计数，免费方案可启动三次并每 30 天增加一次，付费方案需在手机应用中购买。接口与 MCP 共享每分钟 60 次请求的限制。

安装

在 SourceWeft 中

  1. 打开 控制台中的 TGGET Niche Research，将其添加到工作区。
  2. 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": {
    "niche-research": {
      "type": "http",
      "url": "https://tgget.io/mcp"
    }
  }
}

README

TGGET Niche Research MCP server

Remote Model Context Protocol server of TGGET, a niche research tool for mobile apps and online services. It lets an AI assistant check an idea against measured data instead of guessing from a description.

Give it a search phrase or a description of an idea in your own words. It returns monthly search demand, sub-niches, competitor sites from search results, App Store apps with what their users complain about, community signals and a final report with a verdict: worth building, needs validation or no niche. The verdict is computed by rules from five measured factors, so the same numbers always give the same answer; the language model only explains it.

  • Endpoint: https://tgget.io/mcp (Streamable HTTP)
  • Authentication: personal access token, sent as a Bearer token
  • Name in the MCP Registry: io.tgget/niche-research
  • Documentation: tgget.io/docs/api

What you can ask

  • "Validate this idea: an app that reminds you to water plants and explains how to care for them."
  • "Is there demand for an expense tracker for couples? Who are the competitors?"
  • "Research the niche 'invoice generator' and give me the report as Markdown."
  • "Compare my last three researches and tell me which idea to start with."
  • "Follow the niche of my research about habit trackers and show what changed since last month."

Getting a token

  1. Create an account at tgget.io or in the Android app and confirm your email. A free account is enough.
  2. Open Settings, section "API and AI assistants", and create a token. It is shown once.

Tokens expire after a year and are revoked when you change your password or close your sessions. An account holds up to five tokens.

Connecting

Replace <token> with your token. Ready-made files are in examples.

Claude Code

bash
claude mcp add --transport http tgget https://tgget.io/mcp --header "Authorization: Bearer <token>"

Cursor

~/.cursor/mcp.json or .cursor/mcp.json in a project:

json
{  "mcpServers": {    "tgget": {      "url": "https://tgget.io/mcp",      "headers": { "Authorization": "Bearer <token>" }    }  }}

VS Code

.vscode/mcp.json:

json
{  "inputs": [    { "type": "promptString", "id": "tgget-token", "description": "TGGET access token", "password": true }  ],  "servers": {    "tgget": {      "type": "http",      "url": "https://tgget.io/mcp",      "headers": { "Authorization": "Bearer ${input:tgget-token}" }    }  }}

Codex

~/.codex/config.toml, with the token in the TGGET_TOKEN environment variable:

toml
[mcp_servers.tgget]url = "https://tgget.io/mcp"bearer_token_env_var = "TGGET_TOKEN"

Clients that only run local servers

Use the mcp-remote bridge:

json
{  "mcpServers": {    "tgget": {      "command": "npx",      "args": ["-y", "mcp-remote", "https://tgget.io/mcp", "--header", "Authorization: Bearer <token>"]    }  }}

Tools

ToolWhat it doesParameters
start_researchStarts a research from a search phrase or a description of an idea. A phrase is scanned directly; a description is first turned into real search queries and the strongest one is scanned.text (required), deep, locale
get_researchCurrent state and findings: status, headline numbers, ranked candidate phrases, competitor overview, child researches, skipped stages.id (required)
get_reportThe final niche report: verdict, customer pains, MVP, napkin economics, SEO and ASO plans, name options, landing page draft, distribution channels, validation steps and risks.id (required), format (markdown or json)
list_researchThe most recent researches of the account, newest first.limit (1 to 50)
run_signalAdds one check to a finished research: competitors for a phrase, demand history, App Store, reviews, community signals, paid search metrics, or a rebuilt report.id, kind (required), phrase, locale
compare_researchPuts two to five finished researches side by side and marks the better value of every measure. The marks come from the numbers, not from the language model.ids (required)
follow_nicheStarts following the niche of a finished research that has a report: it is scanned anew every month and a digest of what changed is sent to the account. Monthly checks do not spend researches.research_id (required)
list_projectsThe niches the account follows, with the latest numbers and the date of the next check.limit (1 to 50)
get_projectA followed niche with the history of its checks and what changed since the previous one.id (required)
update_projectPauses a followed niche, follows it again or stops following it.id, status (active, paused or deleted) (required)

locale sets the language of the conclusions and the report; en is the one available now. The market is chosen by the language of the phrase: search demand is measured in the country of that language. Texts meant for a site or an app store are always written in the language of the market.

Resources and prompt

NameWhat it gives
tgget://quotaResearches remaining on the plan and when the next one becomes available.
tgget://sourcesResearch kinds available right now and what each one costs.
find_niche (prompt)Walks the assistant through validating an idea: start, wait, read the evidence, write a verdict. Argument: idea.

How a research runs

Research runs in the background and takes a few minutes.

  1. Call start_research. It returns the research id.
  2. Poll get_research every 20 to 30 seconds until status is completed and pending is 0. Findings appear as they arrive.
  3. When the research lists a report, call get_report.

Limits and privacy

  • A research is counted when it starts. Checks added to an existing research with run_signal are free.
  • The free plan includes three researches to start and one more every 30 days. Paid plans are on the plans page and are bought in the mobile app; the same account works everywhere.
  • A free account follows one niche, Starter two, Pro ten.
  • Researches made on the free plan are public: the result may be listed in the open niche catalog. The catalog shows the researched phrase and its findings, never the description you typed. Researches on paid plans stay private.
  • The API and the MCP server share a limit of 60 requests per minute.
  • Search interest is evidence, not a customer count. Treat the verdict as a starting point for your own validation.

Links

This repository holds the documentation and connection examples of the hosted server. The server itself runs at tgget.io.

来源：README.md，提交 531df30

工具

0
工具元数据尚未被收录。

版本历史

1
  1. v1.0.1最新Sep 29, 2026