feat(scraping): add get_saved_posts tool for LinkedIn bookmarks

[blaisebruno]

What this adds

A new get_saved_posts tool that navigates to https://www.linkedin.com/my-items/saved-posts/ and extracts the user's bookmarked posts, including author, full text content, and post URLs.

Why

Saved posts is one of the most-used LinkedIn features for power users (content research, reading lists, CRM workflows). It is the only major user-facing page not yet covered by the server.

Changes

• linkedin_mcp_server/tools/messaging.py: new get_saved_posts MCP tool registered with readOnlyHint, limit param (1-50, default 20), same pattern as get_inbox
• • linkedin_mcp_server/scraping/extractor.py: new get_saved_posts method on LinkedInExtractor, plus two private helpers:
• • _expand_see_more_buttons: clicks see-more buttons to unfold truncated posts before scraping
• • _click_show_more_results: handles pagination by clicking the load-more button

Behaviour

• Accepts a limit parameter (1-50, default 20)
• • Expands truncated posts before scraping
• • Scrolls to load more results up to limit
• • Returns references filtered to post URLs (/feed/update/) and profile URLs (/in/)
• • Authentication error handling identical to get_inbox

Tool table entry (for README)

| get_saved_posts | Get bookmarked posts from the LinkedIn saved items page | working |

[greptile-apps]

Greptile Summary

Adds get_saved_posts, a new MCP tool that navigates to https://www.linkedin.com/my-items/saved-posts/, scrolls for pagination, expands truncated posts, and returns bookmarked post text plus filtered references. The tool registration in messaging.py closely mirrors get_inbox and looks correct; the main concern is in the extractor.

• _expand_see_more_buttons is called before the scroll loop, so posts loaded via scrolling or the "show more results" click remain truncated when content is extracted. Moving the expansion call to after all loading steps resolves this.

Confidence Score: 4/5

Safe to merge after fixing the see-more expansion ordering — all other findings are style or minor robustness suggestions.

One P1 logic defect: _expand_see_more_buttons runs before scrolling loads new content, so posts surfaced by scroll/pagination return truncated text. The tool registration and auth error handling are solid. All other findings are P2.

linkedin_mcp_server/scraping/extractor.py — specifically the ordering of _expand_see_more_buttons relative to the scroll and show-more calls.

Important Files Changed

Filename	Overview
linkedin_mcp_server/scraping/extractor.py	Adds get_saved_posts, _expand_see_more_buttons, and _click_show_more_results; expansion runs before scrolling so newly loaded posts remain truncated (P1), and inline import asyncio is redundant.
linkedin_mcp_server/tools/messaging.py	Registers the get_saved_posts MCP tool with correct annotations, limit validation, auth error handling, and progress reporting — closely follows the get_inbox pattern.

Sequence Diagram

sequenceDiagram
    participant Client
    participant Tool as get_saved_posts (tool)
    participant Extractor as LinkedInExtractor
    participant Page as LinkedIn Page

    Client->>Tool: call(limit=N)
    Tool->>Extractor: get_saved_posts(limit)
    Extractor->>Page: navigate /my-items/saved-posts/
    Page-->>Extractor: page loaded
    Extractor->>Page: _expand_see_more_buttons() ⚠️ before scroll
    Page-->>Extractor: initial posts expanded
    Extractor->>Page: _scroll_main_scrollable_region(attempts=limit//5)
    Page-->>Extractor: new posts loaded (unexpanded)
    Extractor->>Page: _click_show_more_results() once
    Page-->>Extractor: additional posts loaded (unexpanded)
    Extractor->>Page: _extract_root_content([main])
    Page-->>Extractor: raw text + references
    Extractor->>Extractor: strip_noise, build_references, filter /feed/update/ and /in/
    Extractor-->>Tool: url, sections, references
    Tool-->>Client: result

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 2232-2241

Comment:
**See-more expansion runs before loading, leaving new posts truncated**

`_expand_see_more_buttons` is called once before the scroll loop, so it only expands posts that were visible on initial page load. Any posts surfaced by `_scroll_main_scrollable_region` or `_click_show_more_results` will still have their content truncated when `_extract_root_content` runs. Moving the expansion call to after all loading steps is complete ensures it covers all loaded posts.

```suggestion
        # Scroll to load more posts up to the limit
        scrolls = max(1, limit // 5)
        await self._scroll_main_scrollable_region(
            position="bottom", attempts=scrolls, pause_time=0.8
        )

        # Click additional "Show more results" buttons if present
        await self._click_show_more_results()

        # Expand truncated posts by clicking "see more" buttons (after all content is loaded)
        await self._expand_see_more_buttons()
```

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 2279-2280

Comment:
**Redundant `import asyncio` inside method bodies**

`asyncio` is already imported at the top of the file (line 5). The inline `import asyncio` in both `_expand_see_more_buttons` (here) and `_click_show_more_results` (line 2297) are unnecessary. Python deduplicates repeated imports but this is non-idiomatic and clutters the method body.

```suggestion
            await asyncio.sleep(0.5)
```

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 2284-2300

Comment:
**`_click_show_more_results` fires only once**

For `limit` values above ~10, a single "show more results" click is unlikely to surface enough posts — the page typically loads a fixed batch per click. Combined with the scroll loop (which may not trigger a "show more results" button on its own), users requesting 30–50 posts may see fewer results than asked for with no indication of truncation. Consider looping the click until the button disappears or a target count is met, similar to how scrolling uses `attempts`.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "feat(scraping): add get_saved_posts tool..." | Re-trigger Greptile}

[greptile-apps]

See-more expansion runs before loading, leaving new posts truncated

_expand_see_more_buttons is called once before the scroll loop, so it only expands posts that were visible on initial page load. Any posts surfaced by _scroll_main_scrollable_region or _click_show_more_results will still have their content truncated when _extract_root_content runs. Moving the expansion call to after all loading steps is complete ensures it covers all loaded posts.

Suggested change

	await self._expand_see_more_buttons()
	# Scroll to load more posts up to the limit
	scrolls = max(1, limit // 5)
	await self._scroll_main_scrollable_region(
	position="bottom", attempts=scrolls, pause_time=0.8
	)
	# Click additional "Show more results" buttons if present
	await self._click_show_more_results()
	# Scroll to load more posts up to the limit
	scrolls = max(1, limit // 5)
	await self._scroll_main_scrollable_region(
	position="bottom", attempts=scrolls, pause_time=0.8
	)
	# Click additional "Show more results" buttons if present
	await self._click_show_more_results()
	# Expand truncated posts by clicking "see more" buttons (after all content is loaded)
	await self._expand_see_more_buttons()

Prompt To Fix With AI

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 2232-2241

Comment:
**See-more expansion runs before loading, leaving new posts truncated**

`_expand_see_more_buttons` is called once before the scroll loop, so it only expands posts that were visible on initial page load. Any posts surfaced by `_scroll_main_scrollable_region` or `_click_show_more_results` will still have their content truncated when `_extract_root_content` runs. Moving the expansion call to after all loading steps is complete ensures it covers all loaded posts.

```suggestion
        # Scroll to load more posts up to the limit
        scrolls = max(1, limit // 5)
        await self._scroll_main_scrollable_region(
            position="bottom", attempts=scrolls, pause_time=0.8
        )

        # Click additional "Show more results" buttons if present
        await self._click_show_more_results()

        # Expand truncated posts by clicking "see more" buttons (after all content is loaded)
        await self._expand_see_more_buttons()
```

How can I resolve this? If you propose a fix, please make it concise.

[greptile-apps]

Redundant import asyncio inside method bodies

asyncio is already imported at the top of the file (line 5). The inline import asyncio in both _expand_see_more_buttons (here) and _click_show_more_results (line 2297) are unnecessary. Python deduplicates repeated imports but this is non-idiomatic and clutters the method body.

Suggested change

	import asyncio
	await asyncio.sleep(0.5)
	await asyncio.sleep(0.5)

Prompt To Fix With AI

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 2279-2280

Comment:
**Redundant `import asyncio` inside method bodies**

`asyncio` is already imported at the top of the file (line 5). The inline `import asyncio` in both `_expand_see_more_buttons` (here) and `_click_show_more_results` (line 2297) are unnecessary. Python deduplicates repeated imports but this is non-idiomatic and clutters the method body.

```suggestion
            await asyncio.sleep(0.5)
```

How can I resolve this? If you propose a fix, please make it concise.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

[greptile-apps]

_click_show_more_results fires only once

For limit values above ~10, a single "show more results" click is unlikely to surface enough posts — the page typically loads a fixed batch per click. Combined with the scroll loop (which may not trigger a "show more results" button on its own), users requesting 30–50 posts may see fewer results than asked for with no indication of truncation. Consider looping the click until the button disappears or a target count is met, similar to how scrolling uses attempts.

Prompt To Fix With AI

This is a comment left during a code review.
Path: linkedin_mcp_server/scraping/extractor.py
Line: 2284-2300

Comment:
**`_click_show_more_results` fires only once**

For `limit` values above ~10, a single "show more results" click is unlikely to surface enough posts — the page typically loads a fixed batch per click. Combined with the scroll loop (which may not trigger a "show more results" button on its own), users requesting 30–50 posts may see fewer results than asked for with no indication of truncation. Consider looping the click until the button disappears or a target count is met, similar to how scrolling uses `attempts`.

How can I resolve this? If you propose a fix, please make it concise.