How to automate LinkedIn outreach with AI agents

Traditional LinkedIn outreach is a grind: manually search for prospects, visit each profile, craft a personalized note, send the invitation, wait, follow up. An SDR might manage 30-50 personalized touches per day. AI agents connected to LinkedIn via MCP can do the research and drafting in minutes - while keeping the personalization that actually gets responses.

Here's the workflow. Step one: define your target. Tell your AI agent something like "Find VP of Sales at B2B SaaS companies with 50-200 employees in the US." The agent calls search_people with the right filters and gets back a list of matches with names, headlines, and profile URLs.

Step two: qualify and research. For each prospect, the agent calls get_profile to read their full background - current role, previous companies, recent activity, shared connections. It can also check get_company for company context. This is where AI shines: it synthesizes all this context into a mental model of each prospect in seconds.

Step three: personalized outreach. The agent drafts a connection request using send_invitation with a custom note that references something specific - their recent post, a mutual connection, their company's latest funding round. This isn't template-based "I saw your profile and..." spam. The agent genuinely reads and responds to each person's context.

Step four is the cap. Crispy's defaults are 15 invitations per day, 150 messages per day, and 100 engagement actions per day. You can read the full default list on LinkedIn safety. LinkedIn's User Agreement section 8.2 prohibits unauthorized automation. The caps do not authorize activity and do not guarantee against restriction.

Step five: follow up and track. After initial outreach, the agent can monitor responses via list_conversations, send follow-up messages to people who accepted but didn't reply, and track acceptance rates. Some teams build full campaign workflows that run daily check-ins through their AI agent.

The agent does the research and drafts each message from the profile it read. Crispy's default caps still apply, and data rights still follow the privacy pages. The caps do not guarantee against restriction.