I built a Claude MCP that runs my LinkedIn prospecting. one week it booked me 11 meetings, I closed 5
retained text
I'm an AI engineer and I build MimikFlow, so yes, this is my own product and I'm biased. posting because the way I built it with Claude Code might be useful to some of you. what it does: it runs LinkedIn outreach end to end. it finds prospects that match your ICP, sends the connection requests, writes the first DM, follows up, and an AI setter handles the replies and pushes toward a booked call. when a convo gets weird or actually important, it hands it back to a human. the Claude part is the MCP connector. I paste one link into Claude (Settings > Connectors) and Claude becomes the control tower. "give me this week's recap", read the conversations, approve or rephrase pending messages, answer the ones the AI escalated to me, tweak the targeting, go find new prospects. everything goes through the same LinkedIn safety caps as the app, so Claude can't blast 500 invites because I phrased something badly. how Claude Code actually helped, honestly: the architecture is mine. that's my job, I designed every agent in there. but I started by building the whole flow as an n8n workflow, because visual was way easier to reason about. once the global architecture worked in n8n, I asked Claude Code to rewrite it as real code. since then it's been about 8 months of iterating with Claude Code on top, with proper engineering habits (rules files, tests), not just vibe coding. so basically: n8n to think, Claude Code to build, me on the architecture. on my own account, one week it booked me 11 meetings and I closed 5. the product is around $11k MRR now with 80+ subscriptions. it's free to try for 14 days, paid plans after that. link is on my profile, not dropping it here. for anyone who went from an n8n prototype to a Claude Code codebase: what broke for you during the rewrite?
document fetched 2026-11-16 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1xwt9vj/i_built_a_claude_mcp_that_runs_my_linkedin