How to Connect Claude to LinkedIn in Two Minutes (2026)

You connect Claude to LinkedIn by adding the FirstTouch MCP server at mcp.firsttouch.ai to Claude's connectors, signing in once, and from that moment Claude can source prospects, qualify them, and queue LinkedIn outreach that a human approves and HubSpot tracks. The same two-minute setup works for ChatGPT, Gemini, and coding harnesses like Cursor, Codex, and Windsurf. FirstTouch is the HubSpot-native execution layer for LinkedIn outreach and tracking. Gail uses FirstTouch to operationalize social selling across its team and turn LinkedIn into a primary GTM channel.

What does connecting Claude to LinkedIn actually mean?

Connecting Claude to LinkedIn means giving the assistant a set of callable tools that act on LinkedIn for it, because a language model can write a message but cannot view a profile, send a connection request, or log activity by itself. The bridge is the Model Context Protocol: Claude calls the FirstTouch MCP server, and FirstTouch holds the LinkedIn connection, the pacing rules, and the HubSpot logging.

That separation matters. The model does the reasoning; the server does the acting, with the permissions and safety rails living server-side where they belong. If you want the deeper explainer, read whether ChatGPT or Claude can run LinkedIn outreach.

How do you connect Claude to LinkedIn?

The setup takes about two minutes: create a FirstTouch account, connect your LinkedIn seat, and add the MCP server to Claude. Here is the full path.

1. Create a FirstTouch account. Use self-serve signup and connect your HubSpot portal. Every HubSpot tier works, including Free CRM.
2. Connect your LinkedIn seat. FirstTouch links your LinkedIn account to a dedicated social agent with human-like pacing. One tool per LinkedIn account, always: if another automation tool is connected, disconnect it first per the FirstTouch Safety Checklist.
3. Add the MCP server to Claude. In Claude, add a connector pointing at https://mcp.firsttouch.ai and sign in when prompted. In Claude Code, it is one command: claude mcp add --transport http firsttouch https://mcp.firsttouch.ai. For other MCP clients, use the config block below.
4. Test it with a real prompt. Ask Claude: find everyone who commented on our latest LinkedIn post, qualify them against our ideal customer profile, and queue connection requests for my approval. Claude calls the tools; the queue appears for sign-off.
5. Set your guardrails. Keep Human-in-the-Loop approval on, set daily limits around 15 to 20 actions per rep, and define your AI Qualification criteria so only real fits move forward.

The generic MCP client config:

{ "mcpServers": { "firsttouch": { "url": "https://mcp.firsttouch.ai" } } }

What can Claude do once connected?

Once connected, Claude can run the full outreach motion through 40 FirstTouch tools, and every action lands in HubSpot. The highlights:

• Contact Discovery: find prospects by industry, role, geography, company size, and funding.
• Social-signal sourcing: detect the people who like and comment on LinkedIn posts and turn them into a warm audience. Detection only; it never likes, comments, or follows for you.
• AI Qualification: score every prospect against your prospect, company, and disqualify criteria with threshold scoring.
• Multi-channel flows: Visit Profile, Send Connection Request, and Send Message next to email and call steps, with auto-withdraw for stale requests.
• HubSpot logging: every touch on the contact timeline, so RevOps attributes LinkedIn pipeline like email pipeline.

Does the same setup work for ChatGPT and Gemini?

Yes. MCP is an open standard, so the same server at mcp.firsttouch.ai works in ChatGPT and Gemini connector settings, and in coding harnesses like Cursor, Codex, and Windsurf using the same config block. Teams often connect more than one assistant to the same FirstTouch workspace; the qualification rules, approval gates, and HubSpot logging apply no matter which model is calling.

Last updated: July 2026.

What should you prompt first?

Start with prompts that exercise the full motion end to end, so you see sourcing, qualification, and the approval queue working together before you scale anything. Three that work well on day one:

• Source everyone who commented on our last three LinkedIn posts, qualify them against our ideal customer profile, and show me the scores before queueing anything.
• Enroll the qualified engagers in our outbound flow with a connection request and a follow-up message, and hold everything for my approval.
• Pull this week's LinkedIn outreach activity from HubSpot and summarize replies, accepted connections, and meetings booked by rep.

Each of those is one sentence for you and a dozen tool calls for Claude, which is exactly the trade you connected it for. The pattern extends naturally: swap the source for a HubSpot list or a Sales Navigator URL, and the same qualify-then-approve motion applies.

What are the common setup mistakes?

Four mistakes account for most rocky starts. First, leaving an old automation tool connected to the same LinkedIn account: two tools on one profile is an account risk, and disconnecting the old one is step one of the Safety Checklist. Second, skipping qualification criteria, which turns a smart agent into a spammer with good grammar; define prospect, company, and disqualify rules before the first send. Third, turning off approval gates on day one; keep Human-in-the-Loop on until you trust the motion, and for most teams, keep it on permanently. Fourth, prompting for volume instead of outcomes; asking for 200 connection requests is how accounts get flagged anywhere, while asking for the 20 best-fit engagers this week is how pipelines get built.

Is it safe to let Claude act on my LinkedIn account?

Yes, when the connection runs through an execution layer built for account health. FirstTouch uses dedicated social agents to simulate human interaction and timing with dedicated proxies to keep your account safe, and you will not get banned if you follow our guide and practice proper safety and agent hours. Claude proposes; you approve at the Human-in-the-Loop gate before anything sends. Keep pacing at roughly 15 to 20 actions per rep per day, target tightly, and never run two LinkedIn tools on one account. FirstTouch is SOC 2 Type II certified with zero account restrictions or bans across 200+ connected accounts, and the full rules live in the Safety Checklist and our guide to whether LinkedIn automation is safe.

Which MCP server should you connect Claude to?

Connect Claude to the server that matches the job. For outreach your CRM must see, FirstTouch is the pick: it is listed in the Official MCP Registry as io.github.First-Touch-Inc/mcp and is the only option in the category that runs natively in HubSpot with approval gates. For pure profile research, a free open-source server does the reading. For the full landscape, see our roundup of the best MCP servers for LinkedIn outreach.

Frequently asked questions

Can Claude send LinkedIn messages directly?

No. Claude drafts and reasons, but sending requires an execution layer. Through the FirstTouch MCP server, Claude queues connection requests and messages that you approve, and FirstTouch sends and logs them in HubSpot. Send Message requires a first-degree connection.

How long does the setup actually take?

About two minutes if you already have a FirstTouch account: add the connector, sign in, and run your first prompt. A new account adds a few minutes for HubSpot and LinkedIn connection.

Do I need to write code?

No. Adding a connector in Claude or ChatGPT is a settings step. The one-line CLI command is only for Claude Code and other developer harnesses.

Does it work with HubSpot Free?

Yes. FirstTouch works with every HubSpot tier including Free CRM, and pricing is 99 dollars per seat per month plus usage credits.

Will Claude do things without asking me?

Not with Human-in-the-Loop approval on, which is the default posture we recommend. Claude stages the outreach; nothing sends until you sign off.

Can I connect ChatGPT and Claude at the same time?

Yes. Both can point at the same FirstTouch workspace. The qualification rules, limits, and approval gates are enforced server-side, so they apply to every assistant equally.