This server is an MCP toolkit for interacting with the LinkedIn API, with OAuth session management, capability awareness, and explicit dry-run write workflows.
OAuth & session management: auth_start, auth_poll, auth_finish, auth_refresh, auth_set_access_token, auth_set_role_hints, auth_status, auth_clear for handling LinkedIn authentication and token state.
Capability inspection: list_tool_catalog, list_endpoint_manifest, list_capabilities, can_execute_tool to see what tools are available and whether they can run against current permissions.
Raw API helpers: linkedin_get and linkedin_post for arbitrary authenticated GET/POST requests to LinkedIn REST endpoints.
Identity & organization access: whoami, get_member_profile, list_accessible_organizations to resolve the current principal and org ACLs.
Posts: create_post (preview by default), get_post, delete_post, plus list_comments, create_comment for social action interactions.
Reactions: list_reactions, create_reaction, delete_reaction for managing social reactions.
Media uploads: initialize_media_upload and finalize_media_upload for image/video asset workflows, with preview by default.
Advertising: list_ad_accounts, list_campaign_groups, create_campaign_group, list_campaigns, create_campaign, update_campaign, get_ad_analytics for ad account/campaign management and analytics.
Safety-oriented writes: Most mutating tools default to execute=false so you can preview requests before actually calling LinkedIn APIs.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@LinkedIn MCP Server Draft a LinkedIn post about our new product launch and show me the preview."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol (MCP) server that provides a practical, capability-aware LinkedIn API toolset to AI agents such as opencode, Claude, Cursor, and other MCP clients.
Related MCP server: LinkedIn MCP Server
Feature domain | stickerdaniel/linkedin-mcp-server | southleft/linkedin-mcp | linkedin-mcp |
OAuth code flow | ✅ Browser-based auth code exchange | ✅ OAuth 2.0 code flow | ✅ Full |
Write preview mode | ❌ (no dry-run default) | ❌ (no dry-run default) | ✅ All write tools default to |
Ads/campaign management | ❌ Not supported | ❌ Not supported | ✅ |
Media upload workflow | ❌ Not supported | ❌ Not supported | ✅ |
Comment/reaction tools | ✅ Basic comment and reaction support | ✅ Basic comment and reaction support | ✅ |
Direct messaging | ✅ (with session browser) / ⚠️ ToS violation via browser scraping | ✅ (with session browser) / ⚠️ ToS violation via browser scraping | ❌ Not included (requires Sales Navigator/partner approval / |
Current scaffold includes:
OAuth code flow for interactive browser login (auth_start -> auth_finish)
token bootstrap and refresh helpers
per-tool capability matrix in src/linkedin_mcp/tool_catalog.json
raw authenticated helpers for fast endpoint expansion
wrapped tools for profile, org ACLs, posts, comments, reactions, media upload, and ads
You need an MCP-capable client (opencode, Claude Desktop, Cursor, etc).
Open the LinkedIn Developer portal and create a new app.
In app auth settings, add an OAuth redirect URL, for example http://127.0.0.1:8765/callback.
Copy Client ID and Client Secret into your environment.
Under products/scopes, request only what you need now (start with sign-in scopes).
Ensure the redirect URL in app settings exactly matches LINKEDIN_REDIRECT_URI.
Option A - local editable install (recommended for development)
Requirements: Python 3.10+.
pip install -e .
linkedin-mcp-fullOption B - uvx from source folder
Requirements: uv (uvx) and a local clone of this repository.
uvx --from . linkedin-mcp-fullCopy .env.example and set values (or export directly in shell).
Required variables (and usually the only ones you need):
LINKEDIN_CLIENT_ID
LINKEDIN_CLIENT_SECRET
LINKEDIN_REDIRECT_URI (example: http://127.0.0.1:8765/callback)
Example for opencode (~/.config/opencode/opencode.json):
{
"mcp": {
"linkedin": {
"type": "local",
"enabled": true,
"command": ["uvx", "--from", "git+https://github.com/oliverhruby/linkedin-mcp.git", "linkedin-mcp-full"],
"env": {
"LINKEDIN_CLIENT_ID": "{env:LINKEDIN_CLIENT_ID}",
"LINKEDIN_CLIENT_SECRET": "{env:LINKEDIN_CLIENT_SECRET}",
"LINKEDIN_REDIRECT_URI": "{env:LINKEDIN_REDIRECT_URI}"
}
}
}
}
Restart your MCP client after config changes.User prompt | Likely tool call(s) | Expected response |
"Start LinkedIn login" |
| Authorization URL + state + listener details |
"Search for company updates" |
| Organization ACL visibility |
"Create a post preview" |
| Dry-run request payload |
"Create the post now" |
| API write result |
"List my campaigns" |
| Ad campaigns list |
"Get ad analytics" |
| Ad analytics data |
"List my reactions" |
| Reactions for a social action |
"Check account capabilities" |
| Per-tool availability with missing scopes/roles |
Tool | Description | Writes? |
| Build OAuth URL and optional local callback listener | ✅ session |
| Poll pending OAuth callback state | |
| Exchange auth code for access token | ✅ session |
| Refresh access token | ✅ session |
| Manual token bootstrap | ✅ session |
| Set local role hints used for gating | ✅ session |
| Show active session metadata | |
| Clear in-memory session and pending OAuth states | ✅ session |
| Show tool capability metadata | |
| Show endpoint coverage scaffold | |
| Show effective per-tool availability | |
| Check one tool against current capabilities | |
| Raw authenticated GET helper | |
| Raw authenticated POST helper | ✅ |
| Resolve current principal profile endpoint | |
| Fetch member profile | |
| Fetch organization ACL visibility | |
| Create post ( | ✅ |
| Fetch post by URN | |
| Delete post ( | ✅ |
| Fetch comments for social action path | |
| Create comment ( | ✅ |
| Fetch reactions for social action path | |
| Create reaction ( | ✅ |
| Delete reaction ( | ✅ |
| Initialize media upload workflow ( | ✅ |
| Finalize media upload workflow ( | ✅ |
| List ad accounts (approval gated) | |
| List ad campaign groups | |
| Create ad campaign group ( | ✅ |
| List ad campaigns | |
| Create ad campaign ( | ✅ |
| Update ad campaign ( | ✅ |
| Fetch ad analytics (approval gated) |
Write tools are explicit and default to dry-run where feasible.
Capability checks prevent many avoidable permission failures.
API behavior still depends on LinkedIn app approvals and user/org roles.
This is a scaffold-first implementation, not complete LinkedIn API coverage.
LinkedIn API availability depends on app-level approvals and account roles.
Token persistence is currently in-memory; restart requires re-auth or env token.
MIT © Oliver Hrubý
Clear active session from memory.
Inputs: no named parameters. Returns: object
Exchange authorization code for token and activate session.
Inputs: state (required), code, client_secret. Returns: object
Poll pending OAuth state for callback status and auth code availability.
Inputs: state (required). Returns: object
Refresh access token using refresh token from session or argument.
Inputs: client_id, client_secret, refresh_token. Returns: object
Set active LinkedIn access token and optional capability hints.
Inputs: access_token (required), source, expires_at, member_urn, +2 more. Returns: object
Update role hints used by capability gating.
Inputs: role_hints_csv (required). Returns: object
Start OAuth authorization by generating URL and optional localhost callback listener.
Inputs: client_id, scopes_csv, open_browser, redirect_uri, +1 more. Returns: object
Return current session status.
Inputs: no named parameters. Returns: object
Check whether a tool is available under current scope and role hints.
Inputs: tool_name (required). Returns: object
Create ad campaign. Preview by default.
Inputs: path, execute, body_json. Returns: object
Create campaign group. Preview by default.
Inputs: path, execute, body_json. Returns: object
Create comment at provided endpoint path. Preview by default.
Inputs: path (required), body_json (required), execute. Returns: object
Create a simple text post. Preview by default; set execute=true to send.
Inputs: author_urn (required), commentary (required), execute, visibility, +1 more. Returns: object
Create reaction at provided endpoint path. Preview by default.
Inputs: path (required), body_json (required), execute. Returns: object
Delete a post by URN. Preview by default.
Inputs: post_urn (required), execute. Returns: object
Delete reaction using full endpoint path. Preview by default.
Inputs: path (required), execute. Returns: object
Finalize media upload workflow for image/video assets. Preview by default.
Inputs: path (required), body_json (required), execute. Returns: object
Fetch ad analytics with custom query payload.
Inputs: path, query_json. Returns: object
Get member profile using /v2/userinfo or member-specific endpoint if provided.
Inputs: member_urn. Returns: object
Get a post by URN.
Inputs: post_urn (required). Returns: object
Initialize media upload workflow for image/video assets. Preview by default.
Inputs: path (required), body_json (required), execute. Returns: object
Raw authenticated GET helper.
Inputs: path (required), query_json, api_version. Returns: object
Raw authenticated POST helper.
Inputs: path (required), body_json, api_version. Returns: object
List organizations where current user has admin-like roles.
Inputs: path. Returns: object
List ad accounts (approval-gated capability).
Inputs: path. Returns: object
List ad campaign groups.
Inputs: path, query_json. Returns: object
List ad campaigns.
Inputs: path, query_json. Returns: object
Return effective capabilities for current session hints.
Inputs: no named parameters. Returns: object
List comments for a resource path, e.g. /rest/socialActions/{urn}/comments.
Inputs: path (required), query_json. Returns: object
List endpoint coverage scaffold for drift tracking.
Inputs: no named parameters. Returns: object
List reactions for a resource path, e.g. /rest/socialActions/{urn}/reactions.
Inputs: path (required), query_json. Returns: object
List tool capability metadata used by approval-aware clients.
Inputs: no named parameters. Returns: object
Update ad campaign fields. Preview by default.
Inputs: path (required), body_json (required), execute. Returns: object
Resolve current principal using a configurable profile endpoint.
Inputs: path. Returns: object
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 34 tools
Most tools are clearly separated by resource and action, such as list_campaigns and create_comment. However, whoami overlaps with get_member_profile, and the capability/metadata tools (list_tool_catalog, list_capabilities, list_endpoint_manifest, can_execute_tool) have fuzzy boundaries that could cause an agent to pick the wrong one.
The dominant snake_case style is readable and mostly follows a verb_noun pattern like list_campaigns and create_post. But the set mixes conventions with one-word whoami, noun_verb linkedin_get/linkedin_post, and auth_* tools that resemble domain-prefixed verbs or nouns, making the pattern inconsistent.
34 tools is too many for an agent to navigate efficiently, especially with 8 auth tools, 4 capability/metadata tools, and raw HTTP helpers padding the surface. The broad LinkedIn scope justifies some size, but several utility tools could be consolidated or removed.
Core workflows for posts, comments, reactions, media upload, and ad campaigns are represented, and raw linkedin_post can fill some gaps. However, there is no update_post, no delete_comment, and no semantic tool to attach uploaded media to a post, leaving notable lifecycle dead ends.
LinkedIn actions for AI agents: search, messaging, posts and invites, as hosted MCP tools.
LinkedIn for AI agents: search people, read profiles, posts and connections via a connected account
LinkedIn for AI agents: inbox, invitations, Sales Navigator search, posts, jobs, quotas, webhooks.
LinkedIn jobs, companies and people for AI agents, plus messages and invites via your own account.