feat: initial LinkedIn adapter for browserkit

7 tools: get_person_profile, get_company_profile, get_company_posts,
search_people, search_jobs, get_job_details, get_feed.

Uses innerText extraction strategy (resilient to LinkedIn DOM churn)
and walks up from ARIA-labelled action buttons for feed post detection.

Made-with: Cursor

Author: jonzarecki

.github/workflows/ci.yml

@@ -0,0 +1,57 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  test:
+    name: Build & Test
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          path: adapter-linkedin
+
+      - uses: actions/checkout@v4
+        with:
+          repository: browserkit-dev/browserkit
+          path: browserkit
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+
+      - name: Install & build @browserkit/core
+        working-directory: browserkit
+        run: |
+          npm install -g pnpm
+          pnpm install --frozen-lockfile
+          pnpm --filter @browserkit/core build
+
+      - name: Patch @browserkit/core dep path
+        working-directory: adapter-linkedin
+        run: |
+          npm pkg set 'devDependencies.@browserkit/core'='file:../browserkit/packages/core'
+          npm pkg set 'peerDependencies.@browserkit/core'='>=0.1.0'
+
+      - name: Install adapter dependencies
+        working-directory: adapter-linkedin
+        run: npm install
+
+      - name: Install Patchright Chromium
+        working-directory: adapter-linkedin
+        run: npx patchright install chromium --with-deps
+
+      - name: Build
+        working-directory: adapter-linkedin
+        run: npm run build
+
+      # Unit tests only — integration tests require a live authenticated LinkedIn session
+      - name: Test
+        working-directory: adapter-linkedin
+        run: npm test
+        timeout-minutes: 10

.gitignore

@@ -0,0 +1,4 @@
+node_modules/
+dist/
+.DS_Store
+*.storage-state.json

README.md

@@ -0,0 +1,71 @@
+# @browserkit/adapter-linkedin
+
+[LinkedIn](https://www.linkedin.com) adapter for [browserkit](https://github.com/browserkit-dev/browserkit) — exposes LinkedIn as MCP tools over your authenticated local browser session. Runs entirely on your machine; no credentials leave localhost.
+
+## Tools
+
+| Tool | Key inputs | Description |
+|---|---|---|
+| `get_person_profile` | `linkedin_username`, `sections?` | Scrape a person's profile. Optional sections: `experience`, `education`, `interests`, `honors`, `languages`, `contact_info`, `posts` |
+| `get_company_profile` | `company_id`, `sections?` | Scrape a company page. Optional sections: `about`, `jobs`, `people`, `insights` |
+| `get_company_posts` | `company_id`, `count?` | Recent posts from a company page |
+| `search_people` | `keywords`, `count?` | Search LinkedIn people |
+| `search_jobs` | `keywords`, `location?`, `count?` | Search LinkedIn job listings |
+| `get_job_details` | `job_id` | Full details for a specific job posting |
+| `get_feed` | `count?` | Your personalised LinkedIn feed |
+
+Plus auto-registered management tools from the framework: `browser` (health check, screenshot, page state, mode switch, navigate), `close_session`.
+
+## Setup
+
+```bash
+# Install
+pnpm add @browserkit/adapter-linkedin
+
+# One-time login (opens a browser window — sign in normally)
+browserkit login linkedin
+
+# Start
+browserkit start --config browserkit.config.js
+```
+
+```js
+// browserkit.config.js
+import { defineConfig } from "@browserkit/core";
+
+export default defineConfig({
+  adapters: {
+    "@browserkit/adapter-linkedin": {
+      port: 3848,
+      channel: "chrome",   // use real Chrome — avoids bot detection on login
+    },
+  },
+});
+```
+
+Connect your MCP client (Cursor, Claude Desktop, etc.) to `http://127.0.0.1:3848/mcp`.
+
+## How it works
+
+LinkedIn's DOM changes frequently. This adapter uses an **`innerText` extraction strategy** rather than CSS class selectors: it navigates directly to section URLs and reads raw text content. This makes it resilient to React component upgrades and class-name churn.
+
+For the feed tool, post cards are located by walking up from their social action buttons (`aria-label` containing "like"/"comment"/"repost") — stable semantic anchors that don't rotate.
+
+Auth uses a persistent browser profile (`~/Library/Application Support/browserkit/profiles/linkedin` on macOS). Cookies survive daemon restarts — you only need to `browserkit login linkedin` once.
+
+## Session notes
+
+- Use `channel: "chrome"` in your adapter config to use real Google Chrome. This avoids LinkedIn's bot detection on the login page.
+- The adapter runs headless by default. Use the `browser` MCP tool with `action: "set_mode"` and `mode: "watch"` to make it visible for debugging.
+- Rate limit is 3 seconds between tool calls by default to avoid triggering LinkedIn's anti-scraping measures.
+
+## Tests
+
+```bash
+pnpm test                # unit tests (schema, scraper logic, URL building)
+pnpm test:integration    # live browser tests against real LinkedIn (requires login)
+```
+
+## License
+
+MIT

package.json

@@ -0,0 +1,30 @@
+{
+  "name": "@browserkit/adapter-linkedin",
+  "version": "0.1.0",
+  "description": "LinkedIn adapter for browserkit — profile, company, jobs, people search, and feed",
+  "type": "module",
+  "main": "./dist/index.js",
+  "types": "./dist/index.d.ts",
+  "exports": { ".": { "import": "./dist/index.js", "types": "./dist/index.d.ts" } },
+  "scripts": {
+    "build": "tsc",
+    "test": "vitest run",
+    "test:integration": "vitest run --config vitest.integration.config.ts",
+    "lint": "tsc --noEmit"
+  },
+  "peerDependencies": {
+    "@browserkit/core": ">=0.1.0"
+  },
+  "devDependencies": {
+    "@browserkit/core": "github:browserkit-dev/browserkit#main",
+    "@types/node": "^22.13.14",
+    "patchright": "^1.51.1",
+    "tsx": "^4.19.3",
+    "typescript": "^5.7.3",
+    "vitest": "^3.0.9",
+    "zod": "^3.24.2"
+  },
+  "engines": {
+    "node": ">=20"
+  }
+}