From 5906d4ee3ff701dfe754dc625adbec314788bb20 Mon Sep 17 00:00:00 2001
From: Panniantong <panniantong@users.noreply.github.com>
Date: Wed, 25 Feb 2026 13:16:00 +0100
Subject: [PATCH] =?UTF-8?q?fix:=20LinkedIn=20MCP=20=E5=8F=82=E6=95=B0?=
 =?UTF-8?q?=E9=80=82=E9=85=8D=20=E2=80=94=20=E7=94=A8=20company=5Fname/lin?=
 =?UTF-8?q?kedin=5Fusername=20=E6=9B=BF=E4=BB=A3=20url?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

linkedin-scraper-mcp 的 API 参数是 company_name 和 linkedin_username，
不是 url。从 URL 中提取 slug 传给 MCP。
timeout 增加到 60 秒（浏览器自动化需要时间）。
search 去掉不支持的 limit 参数。
---
 agent_reach/channels/linkedin.py | 36 +++++++++++++++++++++-----------
 1 file changed, 24 insertions(+), 12 deletions(-)

diff --git a/agent_reach/channels/linkedin.py b/agent_reach/channels/linkedin.py
index 63d0e29..c9b509f 100644
--- a/agent_reach/channels/linkedin.py
+++ b/agent_reach/channels/linkedin.py
@@ -89,13 +89,19 @@ class LinkedInChannel(Channel):
 
     async def _read_profile_mcp(self, url: str) -> ReadResult:
         """Read a LinkedIn profile via MCP."""
-        safe_url = url.replace('"', '\\"')
+        import re
+        # Extract username from URL: /in/username/
+        match = re.search(r"/in/([^/]+)", url)
+        if not match:
+            return await self._read_jina(url)
+        username = match.group(1)
+        safe_username = username.replace('"', '\\"')
         out = _mcporter_call(
-            f'linkedin.get_person_profile(url: "{safe_url}")',
-            timeout=30,
+            f'linkedin.get_person_profile(linkedin_username: "{safe_username}")',
+            timeout=60,
         )
         return ReadResult(
-            title=self._extract_title(out) or "LinkedIn Profile",
+            title=self._extract_title(out) or f"LinkedIn Profile - {username}",
             content=out.strip(),
             url=url,
             platform="linkedin",
@@ -103,10 +109,16 @@ class LinkedInChannel(Channel):
 
     async def _read_company_mcp(self, url: str) -> ReadResult:
         """Read a LinkedIn company page via MCP."""
-        safe_url = url.replace('"', '\\"')
+        import re
+        # Extract company name from URL: /company/name/
+        match = re.search(r"/company/([^/]+)", url)
+        if not match:
+            return await self._read_jina(url)
+        company = match.group(1)
+        safe_company = company.replace('"', '\\"')
         out = _mcporter_call(
-            f'linkedin.get_company_profile(url: "{safe_url}")',
-            timeout=30,
+            f'linkedin.get_company_profile(company_name: "{safe_company}")',
+            timeout=60,
         )
         return ReadResult(
             title=self._extract_title(out) or "LinkedIn Company",
@@ -202,20 +214,20 @@ class LinkedInChannel(Channel):
         # Try job search first (most common use case)
         try:
             out = _mcporter_call(
-                f'linkedin.search_jobs(keywords: "{safe_q}", limit: {limit})',
-                timeout=30,
+                f'linkedin.search_jobs(keywords: "{safe_q}")',
+                timeout=60,
             )
             results = self._parse_search_results(out, "job")
             if results:
-                return results
+                return results[:limit]
         except Exception:
             pass
 
         # Try people search
         try:
             out = _mcporter_call(
-                f'linkedin.search_people(keywords: "{safe_q}", limit: {limit})',
-                timeout=30,
+                f'linkedin.search_people(keywords: "{safe_q}")',
+                timeout=60,
             )
             results = self._parse_search_results(out, "people")
             if results: