How to Build Your Own LinkedIn AI Coach (No Code, 2026)

Open your LinkedIn analytics right now. Forty charts, and not one of them answers the only question that matters: what should I do today?

That is the dashboard problem. Tools ship metrics because metrics are easy to ship. What you actually want is a coach: something that reads your numbers every morning and hands you three things to do, then shows you the charts only when you ask.

That coach does not exist as a product. This guide shows you how to build it anyway: a small conversational app on your own Taplio data, vibe-coded with Claude in an afternoon, running privately on your machine. You will not write a line of code.

Why a coach beats a dashboard

A dashboard optimizes for completeness. Every metric, every filter, every date range. The cost is that you do the interpreting, every single time, and interpretation is exactly the part you skip when you are busy.

A coach optimizes for the next action. It opens with the three things you need to know today and what to do about them: a checklist of three to five items you can actually clear. No overwhelm, no scroll. When you want the evidence behind the advice, you ask, and the charts appear.

The difference is not the data. It is the same data. The difference is who does the thinking at 8AM.

What vibe-coding means (for the non-developers)

You describe what you want in plain language. Claude writes the code, runs it, hits the errors, and fixes them. You never read a stack trace alone. If you can write a clear prompt, you can ship a working app: the skill that matters is knowing what you want, and you already have that.

What you need

• A Taplio account with your API key. The Taplio API exposes everything Taplio knows about your LinkedIn as clean REST endpoints. Your key is in Taplio under Settings, then Integration. Treat it like a password.
• Claude Code (ships with paid Claude plans), or any Claude chat that can write files to your machine. The docs live at api.taplio.com, OpenAPI spec included.
• An LLM API key for the coach's brain. The app embeds an agent, so Claude will wire in a call to a language model (for example the Claude API) and ask you for a key. It goes in the same local config file as your Taplio key.
• One afternoon. The build itself takes minutes. The afternoon is for playing with it.

Prefer to watch the whole build, from empty folder to first conversation?

The prompt

This is the entire build. One prompt, pasted into Claude Code, in an empty folder.

I want to build a personal LinkedIn growth coach using the Taplio API. The docs are at api.taplio.com (REST v1, API key auth). Build me a small local web app powered by an agent, make it conversational: telling me the three things I need to know every day about my profile and what I need to do to optimize it. The goal is for me not to feel overwhelmed by receiving a checklist of three to five things I need to do every day. When asked, the agent can pull up my analytics, giving three views: my follower and impressions growth over time (GET /v1/analytics/overview), my posts ranked by engagement so I can see which hooks actually work (GET /v1/analytics/posts), and my content pipeline of drafts and scheduled posts (GET /v1/posts and /v1/posts/drafts). One page, clean, runs on localhost. Keep my API key in a local config file and never expose it in anything public.

Nothing to edit. Paste it as is. When Claude asks for your API key, paste it in the local config file it creates, never in a shared conversation.

Here is the whole build on one page, if you would rather save it than scroll back:

What happens when you run it

Step 1: Claude asks for your keys

First your Taplio API key (Settings, then Integration), then a language model key for the agent. Both land in a local config file on your machine. This is the moment that decides whether your build is safe, and Claude handles it correctly when the prompt says so, which this one does.

Step 2: It scaffolds the app

A small local web server, one clean page, a chat interface, and the wiring to the Taplio API. Claude runs it, finds its own errors, and fixes them. Expect a couple of questions about taste, the kind you would get from a contractor: light or dark, chart style, tone of the coach.

Step 3: Open localhost, the briefing is already running

Open localhost. No prompt needed: the briefing runs the moment the page loads. The coach reads your recent numbers and hands you three things you need to know, plus a short checklist of what to do about them. Ask for evidence and it pulls the views live:

• "Show me my growth" calls GET /v1/analytics/overview: your followers and impressions as a daily time-series.
• "Which hooks worked?" calls GET /v1/analytics/posts: every post with its engagement metrics, ranked. The first lines of your winners are your hook playbook.
• "What's in my pipeline?" calls GET /v1/posts and GET /v1/posts/drafts: drafts, scheduled, sent. The coach sees the week ahead, so its advice is about your next post, not your last one.

What's happening under the hood

The Taplio API is a REST API (v1) with API-key auth and a downloadable OpenAPI spec, which is exactly the kind of thing coding agents are good with. The coach uses four endpoint families:

Your growth
GET /v1/analytics/overview returns daily account metrics over any date window.

Your posts
GET /v1/analytics/posts returns per-post engagement, paginated.

Your pipeline
GET /v1/posts and GET /v1/posts/drafts cover everything from draft to sent. The API can also create, schedule and publish drafts, so a coach with a publish button is entirely possible. Your key, your account, your button.

Your niche
GET /v1/inspirations searches the viral-post index programmatically, by topic, engagement thresholds and recency. The coach in this guide does not use it. Your next build might.

One distinction worth keeping straight: this is not the Taplio LinkedIn MCP. The MCP lets AI drive Taplio for you inside a chat. The API lets you build your own tools on top of Taplio. Same data, opposite direction.

Builds worth stealing next

Hook leaderboard. The first lines of your top 20 posts, ranked by engagement. One page, brutal clarity.

Best-time heatmap. Engagement by day and hour posted, from your own history rather than someone's 2019 study.

Growth vs cadence. Your follower curve overlaid on your posting frequency. The chart that settles the "does posting more work" debate with your own data.

Niche radar wall. /v1/inspirations rendered as a live wall of what is rising in your niche.

Pipeline board. Drafts, scheduled, sent as columns, with a publish button wired to the API.

The Monday report. A script that emails you last week's numbers before your coffee. No app, no page, just the habit.

5 tips for a better coach

1. Give it your goals. Tell the agent what you are optimizing for (followers, inbound leads, consistency) and the three daily things get sharper. A coach without a goal is a commentator.

2. Keep the briefing at three items. You will be tempted to ask for more. The moment it becomes ten items, you have rebuilt the dashboard you were escaping.

3. Iterate in plain language. "The tone is too cheerful", "add a weekly view", "remind me when a scheduled slot is empty". Every tweak is one sentence to Claude, not a feature request to a vendor.

4. Let it see your drafts before you post. The pipeline view is the underrated one: a coach that knows what you are about to publish gives advice about the future, which is the only kind you can act on.

5. Restart it weekly, not hourly. The value is the morning ritual, not real-time monitoring. Checking your coach five times a day is the dashboard habit wearing a new coat.

FAQ

Do I need to know how to code?

No. You need to describe what you want and paste an API key when asked. Claude writes, runs and fixes the code. Reading the code afterwards is optional and, honestly, a nice way to learn.

What does it cost?

The Taplio API comes with your Taplio account, and your key is in Settings, then Integration. The agent's brain calls a language model, so bring an LLM API key; usage for one person's daily briefing is small, and you can watch it in your provider's console.

Is my API key safe in this setup?

Locally, yes: the key lives in a config file on your machine and calls go straight from your computer to the Taplio API. The rules are the same as any password: never paste it into a shared chat, never commit it to a public repo.

Can the coach post to LinkedIn without me?

Only if you build it that way. The API does include publish and schedule endpoints, but the coach in this guide is read-only: it looks at analytics and your pipeline. Adding a publish button is your call, and it acts on your account with your key, so wire in a confirmation step if you do.

Does this work with ChatGPT or Cursor instead of Claude?

The Taplio API does not care which agent writes the code: it is a plain REST API with an OpenAPI spec. This guide uses Claude Code because the whole build, including running and debugging, happens in one place.

Why does a 403 or 429 come back sometimes?

403 means your Taplio subscription does not cover that call. 429 means you are hitting the rate limit, so space the calls out; a personal coach polling once per conversation will not get near it.

How is this different from the Taplio LinkedIn MCP?

The Taplio LinkedIn MCP connects Taplio to your AI chat, so Claude or ChatGPT can research, draft and schedule for you. The API is the reverse: it hands you the data so you can build your own tools. Use the MCP for workflows, the API for products, even audience-of-one products like this coach.

Build the thing the roadmap will never ship

No tool will ever ship the exact dashboard, or coach, that fits how you work. That used to be a dead end. Now it is one prompt.

1. Copy your API key: Taplio, Settings, then Integration
2. Open Claude Code in an empty folder
3. Paste the prompt above and watch your first briefing appear

One afternoon. Your own coach, on your own data, on your own machine.