Your pipeline, as tools your assistant can call. LinkedGrow exposes 27 tools over JSON-RPC at its own endpoint. Point Claude or any other MCP client at the URL with your API key, and it can find leads, read the replies that came in overnight, draft an answer, or spin up a new agent, all against your own LinkedIn account.
7-day trial · Cancel before day 7 · Everything included
What is the LinkedGrow LinkedIn MCP server?
It is an endpoint that exposes LinkedGrow as 27 tools over JSON-RPC, so any client speaking the Model Context Protocol can drive your account. An assistant can find leads, read replies, draft answers, create or pause an agent, and schedule posts, using your own API key and your own plan quotas.
The context that matters lives behind a login your AI client cannot reach. You already draft, plan and think inside an assistant. Then anything to do with the people actually replying to you means leaving it, opening a dashboard, copying something out, and pasting it back in with the context stripped off.
Reading a reply and drafting an answer is one thought, and doing it across a dashboard and a chat window turns it into copy, paste, re-explain. The assistant never sees who the person is or what the agent already said to them.
Finding out whether the agent is running, how many leads landed, or whether anyone replied means opening the app and looking. There is no way to just ask, which is exactly the kind of question you want answered without changing context.
Plenty of LinkedIn tools ship a REST API and stop there, which means anyone wanting assistant access writes the glue themselves. A protocol the client already speaks removes the glue entirely.
Plenty of integrations will tell you what happened and nothing more. Reading the pipeline without being able to draft a reply or pause an agent means you still open the dashboard for anything that matters.
I plan the week in Claude and then spend the rest of it clicking through a dashboard to find out what happened...
There is nothing to install. The server is an HTTP endpoint, so the client needs a URL and a key and that is the whole setup.
In Settings, generate a key. It is the same lg_live_ credential the public REST API uses, and it carries the quotas of your plan rather than granting anything on top of them.
Give your MCP client the LinkedGrow endpoint and the key. Claude Desktop, Claude Code and anything else speaking MCP over HTTP all take the same two values, and no local process has to stay running.
Say what you would have clicked. Show me replies from this week, draft an answer to the one from the recruiter, pause the agent while I am on holiday. The client picks the tool and the server does the work.
Three groups, drawn straight from the endpoint. Reads are open, writes are bounded, and nothing here can do something the dashboard would refuse.
list_agents, get_agent_status, create_agent, update_agent, start_agent and pause_agent. An assistant can stand up a new agent for a campaign, watch what it is doing, and stop it, without anyone opening the dashboard.
find_leads, list_leads, get_lead, list_replies, draft_reply, research_prospect and agent_analytics. This is the group that earns its keep: ask for today's replies, get the context behind one lead, and draft an answer in your own voice.
draft_post, save_post, update_post, delete_post, schedule_post, schedule_batch, create_carousel, generate_image, get_voice_profile, list_posts, get_post and list_calendar. Posting did not disappear in v2, it just stopped being the only thing.
Batch calls are capped, plan quotas apply to every tool, and a spend limiter sits in front of anything that costs money. A confused assistant hits a refusal rather than a bill, which is the only design that makes handing over an API key reasonable.
Short answers to what people ask most often.
Write one line about what you sell and we will answer honestly.
Email us directlyMCP is the protocol an AI client uses to call tools that live somewhere else. A LinkedIn MCP server exposes LinkedIn work as those tools, so an assistant can find leads or draft a reply by calling them. LinkedGrow runs one at its own endpoint, against your own account.
27 of them, covering the agent lifecycle, leads and replies, and the content side. You can list and create agents, start or pause one, find and read leads, draft replies, research a prospect, then draft, schedule and publish posts, all from the client you already work in.
No. It is an HTTP endpoint speaking JSON-RPC, so any MCP client points at it with a URL and an API key. There is no package to install, no local process to keep alive and nothing to update when we ship a new tool.
Writes are bounded on purpose. Batch sizes are capped, plan quotas still apply, and a spend limiter sits in front of anything that costs money. An assistant working through the server cannot exceed what you could do yourself in the dashboard.
API access is a Business plan feature, and the MCP endpoint authenticates with the same lg_live_ key as the public REST API. Every tool then respects the quotas of the plan behind that key rather than granting anything extra.
One URL and one key, and the client you already work in can find leads, read replies and draft answers on your LinkedIn account. Everything it does stays inside the quotas of your plan.
7-day trial · Cancel before day 7 · Everything included
