feat(browser): implement persistent browser context for session management by irvingpop · Pull Request #128 · stickerdaniel/linkedin-mcp-server

irvingpop · 2026-01-27T18:19:48Z

Summary

Replaces manual session.json file management with Playwright's persistent browser context for more reliable LinkedIn authentication and session persistence.

Motivation

Re-authenticating was a hassle
LinkedIn security features were triggered more frequently

Changes

Core Implementation

New: PersistentBrowserManager class using launch_persistent_context()
Storage: Sessions now persist in ~/.linkedin-mcp/browser-profile/ directory
Automatic: Cookies and state persist automatically (no manual save/load)

Migration

Automatic migration from legacy session.json on first run
Validates session still works before committing migration
Backs up old session to session.json.backup
Graceful fallback with clear error messages

Configuration

Added --user-data-dir CLI option for custom profile locations
Fixed default path confusion (was incorrectly session.json, now browser-profile)
Updated all documentation (README, AGENTS.md)

Breaking Changes

This is a breaking change (v3.0.0):

Session location changed: ~/.linkedin-mcp/session.json → ~/.linkedin-mcp/browser-profile/
Migration is automatic for existing users
Fallback: Run --get-session to re-authenticate if migration fails

Benefits

More reliable: Cookies persist like a real browser
Less maintenance: No manual save/load cycles
Better Docker support: Standard volume mount pattern
LinkedIn-friendly: Behaves more like normal browser usage (fewer CAPTCHAs)

Testing

✅ All existing tests updated and passing
✅ Migration logic tested with mock legacy sessions
✅ Fresh install flow verified
✅ Session persistence verified across restarts

Verification Checklist

Fresh install: --get-session creates profile
Persistence: No re-authentication needed on restart
Session info: --session-info reports correct status
Clear session: --clear-session removes profile
Migration: Legacy session.json migrates successfully
Documentation: README, AGENTS.md updated

Migration Guide for Users

Existing users (v2.x → v3.0):
Migration is automatic! On first run with v3.0, the server will:

Detect your existing session.json
Migrate cookies to new browser profile
Verify the session still works
Backup old session to session.json.backup

Btw, I'm happy to go with whatever version numbering you want here.

… set

Updated the Dockerfile entry poin. Enhanced main.py with clearer logging and configuration handling, including adjustments for interactive and non-interactive modes. Improved documentation in README.md for command usage and options. Changes include: - Removed unnecessary setup flags in entry point. - Refined logging configuration and output. - Updated README for clarity on command-line options and usage.

Add fast, high-ROI tests covering config, utils, exceptions, error handling, authentication, and MCP tools (mocked). Includes pytest configuration, coverage reporting, and CI test job. - Add pytest.ini with async support and pytest-xdist for parallel tests - Add .coveragerc with 45% threshold and branch coverage - Add tests for config schema, loaders, and singleton pattern - Add tests for exception hierarchy and error handler - Add tests for authentication source detection - Add mocked tests for all 5 MCP tools - Add CI test job running Python 3.14 with coverage

…ropics-claude-code-action-digest chore(deps): update anthropics/claude-code-action digest to 231bd75

….io-astral-sh-uv-latest chore(deps): update ghcr.io/astral-sh/uv:latest docker digest to 143b40f

Updates linkedin-scraper from >=3.1.0 to >=3.1.1 which includes a fix for authentication detection that was causing --get-session to hang indefinitely. The upstream fix (joeyism/linkedin_scraper@55f2305) improves is_logged_in() to handle LinkedIn's A/B tested DOM variants by: - Adding URL-based fallback detection - Checking multiple nav selector patterns - Failing fast on auth blocker URLs Resolves: stickerdaniel#95 Related: joeyism/linkedin_scraper#269

…stickerdaniel#125) Updates linkedin-scraper from >=3.1.0 to >=3.1.1 which includes a fix for authentication detection that was causing --get-session to hang indefinitely. The upstream fix (joeyism/linkedin_scraper@55f2305) improves is_logged_in() to handle LinkedIn's A/B tested DOM variants by: - Adding URL-based fallback detection - Checking multiple nav selector patterns - Failing fast on auth blocker URLs Resolves: stickerdaniel#95 Related: joeyism/linkedin_scraper#269

The base image already provides Python, so uv doesn't install a separate Python version to /opt/python. Remove the reference to fix the Docker build failure.

…l#127) The base image already provides Python, so uv doesn't install a separate Python version to /opt/python. Remove the reference to fix the Docker build failure.

greptile-apps · 2026-01-27T18:24:23Z

Greptile Overview

Greptile Summary

Replaces manual session.json file management with Playwright's persistent browser context (launch_persistent_context) for automatic session persistence. This is a breaking change (v2.3.0 → v3.0.0) that migrates from ~/.linkedin-mcp/session.json to ~/.linkedin-mcp/browser-profile/ directory.

Key Changes:

New PersistentBrowserManager class handles browser lifecycle with automatic state persistence
Automatic migration from legacy session.json on first run with validation and backup
Added --user-data-dir CLI option for custom profile locations
Updated all documentation and tests to reflect new architecture
Proper error handling for profile corruption and concurrent access

Benefits:

More reliable session persistence (cookies persist like a real browser)
Eliminates manual save/load cycles
Better Docker support with standard volume mount pattern
Reduces LinkedIn security triggers (behaves more like normal browser usage)

Migration:
The migration flow automatically detects legacy session.json, validates the session is still active, transfers cookies to the new persistent context, verifies login status, and backs up the old file. Users with existing sessions will be automatically migrated on first run of v3.0.0.

Confidence Score: 4.5/5

Safe to merge with minimal risk - well-tested breaking change with automatic migration
Score reflects solid implementation with comprehensive migration logic, thorough test coverage, and proper documentation updates. Minor uncertainty around BrowserManager.context interface compatibility during migration, but overall architecture is sound and follows Playwright best practices.
Check that linkedin_scraper's BrowserManager has a .context property before merging (used in migration at browser.py:266)

Important Files Changed

Filename	Overview
linkedin_mcp_server/drivers/persistent_browser.py	New class implementing Playwright's persistent browser context with proper error handling for corruption and concurrent access
linkedin_mcp_server/drivers/browser.py	Replaced BrowserManager with PersistentBrowserManager, added migration logic from legacy session.json format
linkedin_mcp_server/authentication.py	Updated to use profile_exists() instead of session_exists(), clear_session now removes directory instead of file
linkedin_mcp_server/cli_main.py	Added automatic migration flow from legacy session.json on startup with user feedback
linkedin_mcp_server/setup.py	Updated to create persistent browser context instead of saving session.json manually
linkedin_mcp_server/config/schema.py	Added user_data_dir config field with validation to ensure parent directory is creatable

Sequence Diagram

sequenceDiagram
    participant User
    participant CLI as cli_main.py
    participant Browser as browser.py
    participant Persistent as PersistentBrowserManager
    participant Playwright

    User->>CLI: Start server
    CLI->>CLI: Check needs_migration()
    alt Legacy session exists
        CLI->>Browser: migrate_from_legacy_session()
        Browser->>Browser: Load legacy BrowserManager
        Browser->>Browser: Extract storage state
        Browser->>Persistent: Create new context
        Browser->>Persistent: Transfer state
        Browser->>Browser: Verify login
        Browser-->>CLI: Migration successful
    end
    
    CLI->>Browser: get_or_create_browser()
    Browser->>Persistent: Initialize with user_data_dir
    Browser->>Persistent: start()
    Persistent->>Playwright: Start playwright
    Persistent->>Playwright: Launch persistent context
    Note over Playwright: State persists automatically
    Playwright-->>Persistent: BrowserContext with Page
    Persistent-->>Browser: PersistentBrowserManager
    
    Browser->>Browser: Navigate to LinkedIn
    Browser->>Browser: Verify authentication
    Browser-->>CLI: Authenticated browser
    
    CLI->>CLI: Start FastMCP server
    Note over CLI: Tools use singleton browser
    
    User->>CLI: Shutdown
    CLI->>Browser: close_browser()
    Browser->>Persistent: close()
    Persistent->>Playwright: Close context
    Persistent->>Playwright: Stop playwright
    Note over Persistent: Session persisted

greptile-apps

_{1 file reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-01-27T18:24:26Z

+        await persistent.start()
+
+        # Copy cookies from old session to new persistent context
+        storage_state = await temp_browser.context.storage_state()


Verify BrowserManager.context property exists - this relies on an undocumented interface from linkedin_scraper

Prompt To Fix With AI

This is a comment left during a code review. Path: linkedin_mcp_server/drivers/browser.py Line: 266:266 Comment: Verify `BrowserManager.context` property exists - this relies on an undocumented interface from `linkedin_scraper` How can I resolve this? If you propose a fix, please make it concise.

…ement Replace manual session.json file management with Playwright's persistent browser context. Sessions now persist automatically in browser profile directory, eliminating need for manual save/load cycles. **Major Changes:** - Add PersistentBrowserManager using launch_persistent_context() - Change session storage: session.json file → browser-profile/ directory - Add automatic migration for existing session.json users - Update configuration with --user-data-dir option - Fix CLI default path (session.json → browser-profile) **Breaking Changes:** - Session location changed from ~/.linkedin-mcp/session.json to ~/.linkedin-mcp/browser-profile/ - Automatic migration provided for existing users - Version bumped to 3.0.0 **Benefits:** - More reliable cookie persistence (behaves like real browser) - No manual save/load cycles needed - Better Docker support with standard volume mount pattern - More LinkedIn-friendly (reduces CAPTCHAggers) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

stickerdaniel · 2026-01-27T19:13:25Z

Hey, can you explain the re-authenticating issues you had? The session management is implemented in the upstream scraper; maybe create an issue there suggesting the use of Playwright's persistent browser context.

irvingpop · 2026-01-27T21:08:54Z

Hey Daniel, thanks for the quick response! Maybe I got caught in weird moment, I was getting bitten by the is_logged_in() issue and kept having to re-authenticate every time and it was getting rather tedious.

But that had me thinking: the session IDs won't last forever, and the is_logged_in() detection is bound to break in the future because it is inherently fragile. So why not make it a little bit easier on myself and others by reusing the same browser session, rather than a fresh one every time?

I do agree this could be more elegantly implemented in the upstream scraper, but I saw that project had a really long queue of unreviewed PRs and plus I wanted to verify this was even the right solution so I implemented here. Totally understand if you'd rather see it upstreamed, and if so I can work on that but it'll be a much more circuitous route.

stickerdaniel · 2026-01-27T21:35:55Z

I see where you're coming from, but I think the upstream PR backlog is mostly stale v2 code. My recent issues there were resolved quite fast.

stickerdaniel · 2026-01-27T21:36:59Z

My main constraint is avoiding the maintenance burden of custom session management within this repository

irvingpop · 2026-01-27T21:38:56Z

Fair point, and totally understandable. If I refactored this such that persistent context stuff went into the scraper library, would you accept a PR to utilize that?

stickerdaniel · 2026-01-27T21:40:35Z

Yes absolutely

irvingpop · 2026-01-28T03:35:53Z

Upstream PR: joeyism/linkedin_scraper#270

stickerdaniel and others added 30 commits July 6, 2025 17:57

docs(readme): update LinkedIn setup instructions

a8052c9

feat(stdout): suppress stdout in MCP stdio mode if --no-setup flag is…

3bbe92e

… set

chore(version): bump version to 1.1.5

41512f1

chore(dxt): update manifest.json version to v1.1.5 [skip ci]

8755670

fix(manifest): add search and recommended jobs features

1ee581a

chore(version): bump version to 1.1.6

bbac67e

chore(dxt): update manifest.json version to v1.1.6 [skip ci]

24a5faf

docs(readme): update LinkedIn cookie instructions

204956e

docs(readme): add HTTP transport mode instructions

901d492

Update README.md

2492ce1

docs(release): clarify LinkedIn cookie setup instructions

b041df7

docs(readme): add star history section

2bde3c7

docs(readme): enhance star history display with dark theme support

bdebb18

docs(readme): update acknowledgements and license section

c04f5b8

docs(readme): add usage warning for LinkedIn scraping

83932a7

docs(readme): simplify star history section markup

6ab121d

docs(readme): clarify license statement

2ddecdc

feat(config): improve transport mode configuration

ce2e0c7

feat(config): add keychain clearing functionality

0ddcd33

docs(readme): document keychain clearing option

16e5d6a

fix(main): version retrieval logic for logging

8736739

docs(readme): update LinkedIn cookie login troubleshooting

6cde44a

chore(version): bump version to 1.1.7

f204145

chore(dxt): update manifest.json version to v1.1.7 [skip ci]

e533c68

refactor(config): remove deprecated debug argument

6284a05

style(main): update server stop message emoji

cb523d2

chore(version): bump version to 1.2.0

e119295

chore(dxt): update manifest.json version to v1.2.0 [skip ci]

0c3a6cc

fix(main): logging and output for claude desktop

1d104b5

stickerdaniel and others added 14 commits January 24, 2026 22:17

chore(deps): update ghcr.io/astral-sh/uv:latest docker digest to 143b40f

381092b

chore(deps): update anthropics/claude-code-action digest to 231bd75

f48d360

Merge pull request stickerdaniel#124 from stickerdaniel/renovate/anth…

d0e2777

…ropics-claude-code-action-digest chore(deps): update anthropics/claude-code-action digest to 231bd75

Merge pull request stickerdaniel#123 from stickerdaniel/renovate/ghcr…

103353c

….io-astral-sh-uv-latest chore(deps): update ghcr.io/astral-sh/uv:latest docker digest to 143b40f

chore(release): bump version to 2.3.1

7182ca9

chore(release): bump version to 2.3.1 (stickerdaniel#126)

1a875e2

chore: update manifest.json and docker-compose.yml to v2.3.1 [skip ci]

94d14ee

fix(docker): remove non-existent /opt/python from chmod

dda0d3e

The base image already provides Python, so uv doesn't install a separate Python version to /opt/python. Remove the reference to fix the Docker build failure.

fix(docker): remove non-existent /opt/python from chmod (stickerdanie…

5174359

…l#127) The base image already provides Python, so uv doesn't install a separate Python version to /opt/python. Remove the reference to fix the Docker build failure.

chore: update manifest.json and docker-compose.yml to v2.3.2 [skip ci]

a184788

greptile-apps Bot reviewed Jan 27, 2026

View reviewed changes

irvingpop force-pushed the main branch from b51da1e to 1478090 Compare January 27, 2026 18:25

address review feedback

3537c34

stickerdaniel marked this pull request as draft January 27, 2026 19:13

stickerdaniel force-pushed the main branch 2 times, most recently from fd80f60 to 7661f43 Compare April 3, 2026 07:59

stickerdaniel force-pushed the main branch from 333ac28 to cd6513f Compare April 17, 2026 09:36

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(browser): implement persistent browser context for session management#128

feat(browser): implement persistent browser context for session management#128
irvingpop wants to merge 407 commits intostickerdaniel:mainfrom
irvingpop:main

irvingpop commented Jan 27, 2026

Uh oh!

greptile-apps Bot commented Jan 27, 2026

Uh oh!

greptile-apps Bot left a comment

Uh oh!

greptile-apps Bot Jan 27, 2026

Uh oh!

irvingpop Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

irvingpop commented Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

irvingpop commented Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

irvingpop commented Jan 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

irvingpop commented Jan 27, 2026

Summary

Motivation

Changes

Core Implementation

Migration

Configuration

Breaking Changes

Benefits

Testing

Verification Checklist

Migration Guide for Users

Uh oh!

greptile-apps Bot commented Jan 27, 2026

Greptile Overview

Greptile Summary

Confidence Score: 4.5/5

Important Files Changed

Sequence Diagram

Uh oh!

greptile-apps Bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Jan 27, 2026

Choose a reason for hiding this comment

Uh oh!

irvingpop Jan 27, 2026

Choose a reason for hiding this comment

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

irvingpop commented Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

irvingpop commented Jan 27, 2026

Uh oh!

stickerdaniel commented Jan 27, 2026

Uh oh!

irvingpop commented Jan 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants