关于 Apex by LeadShark — discovery
Apex by LeadShark 的发现端点是一个免认证 MCP 服务,用于向 AI Agent 介绍完整的 LinkedIn 操作工具集。此轻量服务提供设置说明和入门提示词,本身不执行 LinkedIn 操作。
主要功能
- 01免认证 MCP 发现端点
- 02Apex 概览与相关链接
- 03五步设置说明
- 04可复制的入门提示词
适用场景
- 01了解 Apex 工具集提供的能力
- 02按步骤连接账户
- 03使用提示词开始发现与操作
MCP 相关
通过 Streamable HTTP 提供 about_apex、get_setup_steps 和 first_plays 三种免认证信息工具。它用于发现与握手;LinkedIn 操作由独立的完整认证服务执行。
配置
远程服务无需本地安装。选择你使用的客户端,复制配置或使用一键安装。
https://www.apex.new/mcpclaude mcp add --transport http rzere-linkedin https://www.apex.new/mcp{"mcpServers":{"rzere-linkedin":{"type":"http","url":"https://www.apex.new/mcp"}}}在 Cursor 中打开{"servers":{"rzere-linkedin":{"type":"http","url":"https://www.apex.new/mcp"}}}在 VS Code 中打开{"mcpServers":{"rzere-linkedin":{"type":"http","url":"https://www.apex.new/mcp"}}}{"mcpServers":{"rzere-linkedin":{"serverUrl":"https://www.apex.new/mcp"}}}命令在终端中运行;JSON 粘贴到对应客户端的 MCP 配置后重启。需要授权的服务会在首次连接时提示登录。
工具
3此工具无需参数。
README 原文English
来自项目 READMEREADME 暂无中文译文,以下为英文原文;中文介绍见上方。
linkedin-mcp
Governed LinkedIn hands for AI agents. This is the public face of Apex by LeadShark — an MCP loadout that lets Claude, ChatGPT, or Cursor operate a real LinkedIn account in plain English: find buyers, rank intent, comment, connect, and message. Every action is paced, capped, approval-queued where needed, and fully logged.
This repo contains the discovery endpoint — a minimal, dependency-free MCP server, served live at:
https://www.apex.new/mcp(alias:https://www.apex.new/linkedin-mcp)
Mount it in any MCP host and it completes the handshake instantly — no account, no auth — then briefs your agent on what the full loadout is and how to get it.
The mission
Most LinkedIn tooling automates what you already decided: schedule this post, DM whoever comments. Apex is the layer above that — the intelligence that decides what to automate and where to show up. The activation path is intelligence-first, not automation-first. In five words:
Discover demand. Engage with context.
An agent running Apex works a loop: See → Decide → Act. Find the posts and people where attention already is, figure out who's worth engaging and with what angle, then engage — inside the account's safety budget.
The loadout — 40 tools, See → Decide → Act
The full authenticated server at https://apex.leadshark.io/mcp equips your agent with:
See / discover — where is the demand? (Apex, except discover_lead_magnets)
search · discover_lead_magnets · list_signals · feed · list_recent_posts · list_person_posts · list_post_engagers · get_person_activity
Find people and companies, pull a continuously refreshed radar of prospect-rich lead-magnet posts, list your heat-scored inbound signals, turn one viral post into a warm lead list, and see the rooms a prospect is already in (get_person_activity — who they comment on, what they react to). discover_lead_magnets is available on the Pro+ MCP floor (it doubles as content market-signal); the rest of this group is Apex.
Decide / context — who is worth your time?
enrich · enrich_company · get_lead_activity
Structured profiles for a person or account: role, company, activity, connections. The agent ranks ICP fit and intent and picks the angle — comment, connect, or ignore. Built on enrich, the most-used signal in the loadout.
Act — engagement
comment_on_post · react_to_post · send_connection_request · engage_with_comment · manage_invitations
Show up where the prospect's attention already is — warm them up in public before anything lands in their inbox. Triage inbound connection requests too.
Act — messaging
list_recent_messages · get_messages_with_person · send_message · send_email
Read the inbox, pull a full thread with one person, send the DM — deduped, paced, and logged. send_email sends through your own connected Resend account (BYO key).
Company Pages
list_companies (Pro+) · comment_as_company · react_as_company · manage_company_reply_automations
List and schedule to pages you administer on Pro+; acting in the page's voice — commenting, reacting, and running reply-only comment automations as the page — is Apex.
Pages & captured leads (Apex)
list_lead_pages · get_page_leads · list_lead_magnet_links
Read the leads your lead magnets captured — pages with capture stats, the deduped enriched email leads for a page, and your standalone leads.sh capture links — to feed the enrich → act loop.
Automations & scheduled posts (Pro+ baseline)
create_automation · list_automations · get_automation · edit_automation · suggest_automation_settings · schedule_post_with_automation · list_scheduled_posts · get_scheduled_post · edit_scheduled_post · cancel_scheduled_post
The execution layer: comment→DM lead-magnet automations and pre-automated post scheduling. Apex decides what to run; these run it.
Safety, limits & history
manage_activity_limits · set_daily_dm_limit · list_actions
Check budgets before acting, adjust caps, and read the audit trail — what the agent already did, what's staged for approval, why something was refused, and the current safety mode.
Full per-tool reference (parameters, access, tiers): apex.new/tools.md
The safety model (this is the point)
Every write the agent makes shares the same daily/weekly budgets as the account's automations (DMs, replies, connections), honors the do-not-engage list, and is paced and rate-limit-aware. Actions can be refused or queued for the user's approval — and the agent is told why. Limits are calibrated from LeadShark's production LinkedIn volume, not guesswork. Agents are expected to check list_actions before acting and to work within limits rather than around them.
What this repo's server does
This discovery stub speaks MCP over Streamable HTTP (JSON-RPC 2.0 over POST, JSON responses, stateless, no SSE) and exposes three no-auth informational tools:
| Tool | What it returns |
|---|---|
about_apex | What Apex is, tier pricing, trial paths, key links |
get_setup_steps | The 5-step connect flow (account → LinkedIn → 24h unlock → mount MCP → first play) |
first_plays | Copy-paste starter prompts: verify, discover, enrich, act |
It has no LinkedIn powers of its own — it's the handshake and the briefing, not the hands.
Try it
Add https://www.apex.new/mcp as a custom connector / MCP server in your AI app, or run this repo locally:
node server.mjs # Node 18+, zero dependencies, listens on PORT (default 3000)
Handshake by hand:
curl -s -X POST http://localhost:3000/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"curl","version":"0"}}}'
Get the full loadout
- Create a LeadShark account — free 7-day trial, no card.
- Connect the LinkedIn account the agent will operate.
- Unlock the one-time 24-hour Apex operator window (Settings → Apex Settings).
- Add
https://apex.leadshark.io/mcpto your AI app and authorize with LeadShark. - Ask: "Check if you can see LeadShark Apex and list the tools available."
Apex is $119/mo after the trial. LeadShark Pro ($39/mo) automates your lead magnet; Pro+ ($59/mo) scales automations and gated funnels; Apex is the operator loadout on top.
- Website: apex.new · Pricing: apex.new/pricing
- MCP setup guide: apex.leadshark.io/docs/mcp
- Agent operating guide: apex.new/agents.md
- Machine-readable index: apex.new/llms.txt
License
MIT
评论
一个开放协议,规定了 AI 应用如何以统一方式连接外部工具与数据源。客户端负责与模型对话,MCP 服务负责提供工具、资源和提示模板,两者通过标准消息通信,同一个服务可以被不同客户端复用。