# Apex by LeadShark

> Apex gives your AI agent governed LinkedIn hands. Connect Claude, ChatGPT, or Cursor to LeadShark over the Model Context Protocol (MCP) and run your LinkedIn motion in plain English — find buyers, rank intent, comment, connect, and message. Every action is paced, capped, and fully logged, within your limits.

Apex is the agent layer on top of [LeadShark](https://www.leadshark.io). A human assembles it in five steps (about five minutes); an AI agent then operates a real LinkedIn account through the Apex MCP server. Agent actions share the same daily/weekly safety budgets and do-not-engage rules as LeadShark's automations — you build it, you own it, it works within your limits.

LeadShark's caps are calibrated from production LinkedIn volume: 2.5M+ comments processed, 2.8M+ comment replies sent, 2.2M+ DMs delivered, 1M+ connections sent & accepted, 19K+ posts automated. Limits are set from what the infrastructure has actually run — not guesswork.

## Getting started
- [Recipes](https://www.apex.new/recipes): Repeatable LinkedIn plays with paste-ready agent prompts.
- [Apex docs](https://apex.leadshark.io/docs/apex): What Apex is and how to use it.
- [MCP setup guide](https://apex.leadshark.io/docs/mcp): Add the Apex MCP server to Claude, ChatGPT, or Cursor.
- [MCP server endpoint](https://apex.leadshark.io/mcp): The MCP URL an agent connects to.
- [MCP discovery endpoint](https://www.apex.new/mcp): No-auth handshake + info tools (about Apex, setup steps, first plays). Mount it to test MCP without an account; it points you to the real server above. Also at https://www.apex.new/linkedin-mcp. Source: https://github.com/rzere/linkedin-mcp
- [Start a free trial](https://apex.leadshark.io): Create a LeadShark account (7-day trial; 24-hour Apex unlock).

## How an agent connects
1. Create a LeadShark account.
2. Connect the LinkedIn account the agent will operate.
3. Unlock the 24-hour Apex trial in Settings → Apex Settings.
4. Add the Apex MCP server (URL above) to your AI app and authorize with LeadShark.
5. Ask in plain English to run a "play".

## Your first plays (run these right after connecting)
Apex is for finding what's working, replicating it, and building your own plays. The activation path is intelligence-first, not automation-first. In five words: **discover demand, engage with context.** Each play is one sentence the user types to their assistant.

0. **Connect + verify (start here).** "Check if you can see LeadShark Apex and list the tools available." The trust handshake — the user sees their AI now has a real LinkedIn operating layer. Do this before anything else.
1. **Discover what's working.** "Pull 10 top-performing lead-magnet posts and 10 of my latest posts. Compare them, find the pattern gap, and write my next lead-magnet post." Market intelligence → creator strategy → next post.
2. **Enrich the opportunity.** "Enrich the authors and best commenters from those posts. Tell me who's ICP-fit and what angle I should use with each."
3. **Be where the prospect already is.** "Get this prospect's recent activity — who they commented on, what they reacted to, and which posts I should engage with so they actually see me." Comment where their attention already is, before a DM.

Framing: **LeadShark Pro automates your lead magnet. LeadShark Apex discovers demand, understands prospects, and acts where attention already is** — the intelligence that decides what to automate and where to show up.

## What an agent can do (35 MCP tools)
Full details for every tool — parameters, access (read/write), and tier — are in the [Apex tools reference](https://www.apex.new/tools.md).

- See / discover: `search`, `discover_lead_magnets`, `list_signals`, `feed`, `list_recent_posts`, `list_person_posts`, `list_post_engagers`, `get_person_activity`
- Decide / context: `enrich`, `enrich_company`, `get_lead_activity`
- Act — engagement: `comment_on_post`, `react_to_post`, `send_connection_request`, `engage_with_comment`, `manage_invitations`
- Act — messaging: `list_recent_messages`, `get_messages_with_person`, `send_message`
- Company Pages (Pro+): `list_companies`, `comment_as_company`, `react_as_company`
- Automations: `create_automation`, `list_automations`, `get_automation`, `edit_automation`, `suggest_automation_settings`
- Scheduled posts: `schedule_post_with_automation`, `list_scheduled_posts`, `get_scheduled_post`, `edit_scheduled_post`, `cancel_scheduled_post`
- Safety, limits & history: `manage_activity_limits`, `set_daily_dm_limit`, `list_actions`

## Safety model
Every write an agent makes shares the user's daily/weekly budgets (DMs, replies, connections), honors the do-not-engage list, and is paced and rate-limit-aware to protect the account. Actions may be refused or queued for the user's approval, and the agent is told why. Agents should call `list_actions` to dedupe before acting and must respect limits rather than routing around them.

## Optional
- [Apex tools reference](https://www.apex.new/tools.md): Every MCP tool explained in detail.
- [Agent operating guide](https://www.apex.new/agents.md): Conventions and etiquette for agents using Apex.
- [LeadShark](https://www.leadshark.io): The platform Apex is built on.