独立开发者 · 工具发现实战手册 Indie Dev · Tool Discovery Playbook

让用户主动找到你,让 AI 直接调用你 Get Found by Users, Get Called by AI

5 条渠道 · 每条都有可执行清单 · 点击展开实战细节 5 channels · actionable checklists · click to expand details
▸ 全局总览 — 点击卡片跳转详细手册 ▸ Overview — click cards to jump to details
流量 ↓ Traffic ↓
Your Website
navigator.modelContext.registerTool({ ... })
↓ 调用 ↓ Calls
Claude Chrome Extension
浏览器扩展捕获 registerTool Browser ext captures registerTool
Chrome Native
实验阶段 · 原生集成 Experimental · native integration
Local Relay
Claude Desktop · Cursor · VS Code Claude Desktop · Cursor · VS Code

01
Channel 01 · 长尾稳定流量 Channel 01 · Steady Long-tail Traffic
SEO、目录、社区口碑 SEO, Directories & Community
核心战术 Core Tactics
TACTIC 01
关键词策略:按意图分层 Keyword Strategy: Layer by Intent
不要只抢高竞争大词,按搜索意图分层布局:

信息型「什么是 MCP 工具」— 写博客建权威,排名慢但积累品牌
对比型「X vs Y 哪个好」— 做对比页,截获决策阶段用户,转化最高
品牌型「你的工具名 + 教程」— 提前布局,防竞品截流

长尾词(3-5 词的具体问题)竞争低、意图精准、转化高,优先起步。
Don't just chase high-competition keywords. Layer by search intent:

Informational "what is MCP tool" — blog posts build authority slowly
Comparison "X vs Y" — comparison pages capture decision-stage users, highest conversion
Branded "your tool name + tutorial" — claim early, prevent competitors from hijacking

Start with long-tail keywords (3-5 word specific queries) — low competition, precise intent, high conversion.
SEO
TACTIC 02
技术博客 = 最强 SEO 资产 Technical Blog = Your Strongest SEO Asset
每篇文章解决一个具体问题,结尾自然引出工具。

选题公式:「How to [解决用户痛点] using [你的工具]」
例:「How to expose your SaaS tools to Claude in 5 minutes」

内容结构:痛点描述 → 解决方案 → 代码示例 → 工具链接
发布渠道:自己博客(权重归你)+ dev.to + Medium(额外曝光)
Each article solves one specific problem. Naturally introduce your tool at the end.

Topic formula: "How to [solve user pain] using [your tool]"
e.g.: "How to expose your SaaS tools to Claude in 5 minutes"

Content structure: Pain → Solution → Code example → Tool link
Channels: Your blog (owns the SEO weight) + dev.to + Medium (extra exposure)
内容营销Content
TACTIC 03
工具目录提交(不可跳过) Directory Submission (Don't Skip)
Product Hunt — 发布日冲榜可获数千访问,提前 2 周预热
GitHub Awesome 列表 — awesome-mcp-servers、awesome-ai-tools,提 PR
npm / PyPI — 包名即 SEO,开发者直接搜到
AlternativeTo — 截获竞品用户搜索流量
There's an AI for That — AI 工具垂类流量最大聚合平台
Toolify.ai / Futurepedia — 补充覆盖
Product Hunt — Launch day can bring thousands of visits, warm up 2 weeks early
GitHub Awesome lists — awesome-mcp-servers, awesome-ai-tools, submit PRs
npm / PyPI — Package name is SEO, devs search directly
AlternativeTo — Capture competitor search traffic
There's an AI for That — Largest AI tool aggregator
Toolify.ai / Futurepedia — Additional coverage
目录Directories
TACTIC 04
社区运营:先给后取 Community: Give Before You Take
在 HackerNews、Reddit、Discord AI 社群先回答别人问题,建立信任。

当你分享工具时,社区会接受而非反感。

Show HN 格式:「Show HN: [工具名] – [一句话功能]」
一条好的 Show HN 可带来数千精准开发者访问。
Answer questions first in HackerNews, Reddit, Discord AI communities. Build trust.

When you share your tool, the community will welcome rather than resist it.

Show HN format: "Show HN: [Tool] – [One-line description]"
A good Show HN can bring thousands of targeted developer visits.
社区Community
执行清单 Action Checklist
SEO & 目录 · 行动清单 SEO & Directories · Action Items
0 / 10
02
Channel 02 · 最强爆发力 Channel 02 · Highest Reach
短视频 & 信息流 Short Video & Social Feed
各平台打法 Platform-Specific Tactics
平台Platform
核心打法Core Tactic
适合什么Best For
Twitter / X
60s Demo GIF/视频 + 第一句话就是结果 + 正文放链接 60s demo GIF/video + lead with result + link in body text
开发者/技术圈扩散,最强 Dev 流量 Dev community spread, strongest dev traffic
小红书Xiaohongshu
首图必须是结果截图,口语化心得,回复每条评论 Lead image must be the result screenshot, casual tone, reply to every comment
中文用户种草转化,效率工具类表现好 Chinese market, productivity tools perform well
YouTube / B站
标题含工具名+场景词,描述框第一行放链接 Title includes tool name + scenario, first line of description is the link
长效搜索流量,教程类 5-15 分钟最佳 Long-term search traffic, tutorials 5-15 min work best
TikTok / 抖音Douyin
纯结果展示,不解释原理,加文字说明(静音场景) Pure result showcase, no explanations, add text overlays (silent viewing)
大众用户,爆发性强,转化需配合 bio 链接 Mass audience, explosive reach, conversion needs bio link
Newsletter
直接邮件作者,一句话介绍 + Demo 链接 Email the author directly, one-line intro + demo link
精准高质量用户,转化率比普通流量高 3-5 倍 Precise high-quality users, 3-5x higher conversion than average traffic
高转化 Demo 视频公式 High-Conversion Demo Video Formula
视频结构模板 Video Structure Template
0-3s 直接展示最惊艳的结果画面(不要开场 logo)
3-8s 「只需要 X 步 / X 秒」建立惊喜感
8-30s 实时演示核心操作(不剪辑,保真实感)
30-45s 展示第二个使用场景,拓展想象空间
结尾 明确 CTA:「链接在 bio / 评论置顶」
0-3s Show the most impressive result immediately (no intro logo)
3-8s "Just X steps / X seconds" — build the wow factor
8-30s Live demo of core workflow (no cuts, keep it real)
30-45s Show a second use case, expand imagination
End Clear CTA: "Link in bio / pinned comment"
全程加字幕(所有平台 60%+ 用户静音)· 发布后 1 小时内回复前 10 条评论 Always add subtitles (60%+ users watch muted) · Reply to first 10 comments within 1 hour of posting
执行清单 Action Checklist
短视频 & 社交 · 行动清单 Video & Social · Action Items
0 / 8
03
Channel 03 · 增速最快 Channel 03 · Fastest Growing
AI 辅助发现 & AIO 优化 AI-Assisted Discovery & AIO
原理:LLM 怎么「知道」推荐你? How LLMs "Know" to Recommend You
两个来源 Two Sources
① 训练数据 — 博客 / GitHub / Stack Overflow 中提到你的内容 → 权重积累 → LLM 主动推荐
② 实时检索 — Perplexity / SearchGPT 实时抓取 → 页面结构清晰 + 权威性高 → 被引用
① Training data — Blog / GitHub / Stack Overflow content mentioning you → weight accumulation → LLM recommends proactively
② Real-time retrieval — Perplexity / SearchGPT crawls in real time → clear page structure + high authority → gets cited
核心原则:让 AI 一眼看懂你是什么、解决什么问题、什么时候用你 Core principle: Let AI instantly understand what you are, what you solve, when to use you
核心战术 Core Tactics
TACTIC 01
写给 AI 看的页面内容 Write Page Content for AI
首页加一句话定义:
「[工具名] 是 [类别] 工具,帮助 [用户] 解决 [问题]」

文档加 FAQ 页面,用 Q&A 格式,问题要和用户真正会问 AI 的一致。
FAQ 被 AI 搜索引用的概率比普通段落高 3 倍
Add a one-line definition to your homepage:
"[Tool] is a [category] tool that helps [users] solve [problem]"

Add an FAQ page to your docs using Q&A format. Questions should match what users actually ask AI.
FAQ pages are 3x more likely to be cited by AI search engines
AIO
TACTIC 02
在 AI 训练数据源布点 Place Markers in AI Training Sources
GitHub README — 结构化 + 徽章 + 截图 + 快速上手代码,LLM 训练数据的高权重来源
dev.to / Medium — 被 LLM 训练数据广泛收录
Stack Overflow — 在相关问题下给出包含你工具的解决方案
HN Show HN — 被几乎所有主流 LLM 数据集收录
GitHub README — Structured + badges + screenshots + quickstart code, high-weight LLM training source
dev.to / Medium — Widely included in LLM training data
Stack Overflow — Post solutions that include your tool on relevant questions
HN Show HN — Included in nearly all major LLM datasets
训练数据Training Data
自测:AI 可见度检查(每月一次) Self-Test: AI Visibility Check (Monthly)
AI 可见度 · 行动清单 AI Visibility · Action Items
0 / 8
04
Channel 04 · 技术接入 Channel 04 · Technical Integration
AI Agent 调用接入 AI Agent Integration
第一步:在网页注册工具(通用基础) Step 1: Register Tools on Your Page (Foundation)
// Register your tool on the web page — shared foundation for all 3 channels if ('modelContext' in navigator) { navigator.modelContext.registerTool({ name: 'search_products', // description is how AI decides WHEN to call you // Formula: [what it does] + Use this when [trigger] + Returns [output] description: `Search the product catalog. Use this when the user wants to find products, check availability, or compare items. Returns product name, price, and stock status.`, parameters: { type: 'object', properties: { query: { type: 'string', description: 'Natural language search query', examples: ['red shoes', 'wireless headphones under $100'] }, limit: { type: 'number', default: 10 } }, required: ['query'] }, handler: async ({ query, limit = 10 }) => { const data = await fetchProducts(query, limit); return { products: data, total: data.length }; } }); }
三种接入通道对比 Three Integration Channels Compared
通道Channel
接入方式How It Works
适用场景Best For
Claude Chrome Extension
浏览器扩展注入,捕获 registerTool 声明,暴露给浏览器内 AI agents Browser extension injects, captures registerTool declarations, exposes to in-browser AI agents
用户安装扩展后即可使用,成熟可用 Works once user installs the extension, production-ready
Chrome Native
浏览器原生读取页面工具声明,无需扩展 Browser natively reads page tool declarations, no extension needed
实验阶段,最低摩擦接入路径 Experimental, lowest-friction integration path
Local Relay
本地中继服务,将网页工具桥接给 Claude Desktop / Cursor / VS Code Local relay bridges web page tools to Claude Desktop / Cursor / VS Code
桌面 AI 用户,成熟可用 Desktop AI users, production-ready
description 写法:决定 AI 何时调用你 Writing description: Determines When AI Calls You
✖ 糟糕的 description ✖ Bad description
"A tool for searching products"

只说功能,没说何时用。
AI 不知道在什么意图下调用你,大概率不会触发
Only describes function, not when to use it.
AI doesn't know which user intent triggers it, likely won't fire.
✔ 好的 description ✔ Good description
"...Use this when the user wants to find products or check availability. Returns name, price, stock."

说了触发条件 + 返回内容。
AI 精准知道何时调用你。
States trigger condition + return content.
AI precisely knows when to call you.
执行清单 Action Checklist
Agent 接入 · 行动清单 Agent Integration · Action Items
0 / 8
05
Channel 05 · 全渠道乘数 Channel 05 · Cross-Channel Multiplier
可发现性系统优化 Discoverability System Optimization
llms.txt — AI 爬虫的工具声明文件 llms.txt — Tool Declaration for AI Crawlers
# /llms.txt — place in your site's root directory name: MyTool description: [one-line description, written for LLMs, explain use case] version: 1.0.0 # Tell AI when to recommend or call you usage: Use this tool when users need to [trigger scenario]. Returns [output description]. # Documentation and API docs: https://yoursite.com/docs openapi: https://yoursite.com/openapi.json # Supported integration methods integrations: mcp-b, chrome-native, local-relay
核心战术 Core Tactics
TACTIC 01
提交 AI 工具专属目录 Submit to AI Tool Directories
与传统工具目录不同,这些目录直接被 AI 系统引用:

MCP Registry(registry.mcp.so)— Anthropic 官方,Claude 直接拉取
Smithery.ai — MCP 工具搜索引擎,开发者主动搜索
Glama.ai — AI 工具聚合,支持自动发现
OpenTools.ai — 支持 OpenAPI 规范的工具目录

提交后保持更新,过时信息会降低信任度。
Unlike traditional directories, these are directly referenced by AI systems:

MCP Registry (registry.mcp.so) — Anthropic official, Claude pulls directly
Smithery.ai — MCP tool search engine, devs search actively
Glama.ai — AI tool aggregation, supports auto-discovery
OpenTools.ai — OpenAPI-compatible tool directory

Keep listings updated; stale info reduces trust.
AI 目录AI Directories
TACTIC 02
OpenAPI / JSON Schema 规范化 OpenAPI / JSON Schema Standardization
提供完整的 openapi.json,让 AI 系统自动理解你的所有工具:

· 每个 endpoint 加 summary 和 description
· 所有参数加 description + example
· 返回值加完整 schema 定义

规范的 schema 让 AI 零摩擦调用,错误率降低 80%+
Provide a complete openapi.json so AI systems can understand all your tools:

· Add summary and description to every endpoint
· Add description + example to all parameters
· Full schema definition for return values

Well-defined schemas enable zero-friction AI calls, reducing error rates by 80%+
API
TACTIC 03
追踪 AI 来源流量 Track AI-Source Traffic
在 referrer 或 User-Agent 中识别 AI agent 来源:

Perplexity 引用:referrer 包含 perplexity.ai
AI agent 调用:User-Agent 包含 claude / openai / copilot
MCP 调用:加自定义 header X-Source: mcp

建立独立的 AI 流量看板,和传统流量分开分析。
Identify AI agent sources in referrer or User-Agent:

Perplexity citations: referrer contains perplexity.ai
AI agent calls: User-Agent contains claude / openai / copilot
MCP calls: add custom header X-Source: mcp

Build a separate AI traffic dashboard, analyze independently from traditional traffic.
追踪Tracking
TACTIC 04
稳定性 = 可发现性 Stability = Discoverability
工具不稳定会被 AI 系统降权推荐:

· 版本控制:提供 changelog,旧版本保持兼容
· 错误处理:返回清晰的错误信息,AI 可以理解
· 响应时间:尽量 < 500ms,AI agent 有超时机制
· 状态页面:提供 /status 接口,AI 检查可用性
Unstable tools get deprioritized by AI systems:

· Versioning: Provide changelog, keep old versions compatible
· Error handling: Clear error messages AI can understand
· Response time: Target < 500ms, AI agents have timeout limits
· Status page: Provide /status endpoint, AI checks availability
可靠性Reliability
执行清单 Action Checklist
系统优化 · 行动清单 System Optimization · Action Items
0 / 10