Connect your AI Agent to B2B Channels. Powered by LinkupAPI — a unified API that lets agents act on the channels where B2B revenue actually happens. Today: full LinkedIn coverage (outbound, engagement, content, recruiter, enrichment, webhooks). More channels are landing on the same action-based interface.
This repo is a bundle of Claude skills for LinkedIn: 4 ready-to-run, auditable Claude Code workflows that turn LinkupAPI's capabilities into safe, rate-limit-aware automation. The fastest way to put an AI agent on a real LinkedIn account without writing glue code.
Keywords: claude skills for linkedin · claude code linkedin · linkedin automation claude · claude agent linkedin · linkupapi mcp · linkedin outreach AI agent.
Define an ICP, find target companies, extract decision makers (CEO/CTO/Founder), filter false positives, send connection requests, optionally arm a webhook for accept-triggered messages. 8-stage flow with intent-first ICP discovery (sell / recruit / network), draft ICP proposal, batched-form filter editing, multi-account selection with dedup-and-round-robin, and mandatory campaign log persistence.
Given a LinkedIn post URL, scrape commenters and reactors, filter on visible signal (no enrichment), enrich the matches, score against the ICP, and launch a targeted outreach campaign on confirmed matches. The "saw your comment on…" note boosts accept rate substantially over cold.
Scroll the feed, identify posts authored by ICP-matching profiles, generate value-adding 2-3 sentence comments (auto, templated, or HITL-approved), and post them with strict pacing (30-60s between, max 15/day). Used as a warm-up before outreach: prospect sees your name on their post, then receives an invite.
Given N LinkedIn profile URLs, return enriched data: current role, company, location, experience, skills, plus professional email when findable. Output as inline markdown table, CSV, or JSON. De-dups against ./enrichments/ from the last 30 days so you never re-pay for the same lookup.
Each skill is rate-limit-aware and respects these per-account daily caps before any tool call:
| Cap | Limit / day | Tool action it gates |
|---|---|---|
| Connection requests | 20 | linkedin_network/invite |
| Profile gets | 100 | linkedin_profiles/get |
| Searches | 15 | linkedin_profiles/search_* |
| Comments (feed-engage only) | 15 | linkedin_content/comment |
The cap is shared across all 4 skills — hitting it from any one carries over. Each skill queries linkupapi_get_logs at Stage 0 to compute remaining budget, then trims the campaign size or stops gracefully if the budget is exhausted. Same-day duplicate work is deduped via ./campaigns/*.md and ./enrichments/*.md logs.
git clone https://github.com/claude-dev-code/claude-skills-linkedin /tmp/claude-skills-linkedin
cp -R /tmp/claude-skills-linkedin/linkedin-* ~/.claude/skills/Restart Claude Code. The skills are auto-discovered and triggered by natural-language prompts: "outreach campaign", "enrich these profiles", "comment on my feed", "find leads from this post URL", etc.
Each skill is also published as a .zip artifact on the latest GitHub release. To install:
- Download the zip(s) you want — one per skill:
linkedin-outreach.zip,linkedin-high-intent.zip,linkedin-feed-engage.zip,linkedin-enrich.zip. - In Claude Desktop or claude.ai, go to Settings → Skills → Upload skill and drop the zip.
- The skill becomes available across all your projects/conversations.
The MCP setup is the same on all three surfaces — see below.
All 4 skills require the LinkupAPI MCP server with at least one channel account in connected status. The auth is OAuth — no API key in headers, no token to manage.
Claude Code:
claude mcp add --transport http linkupapi https://mcp.linkupapi.com/mcpThen run /mcp and follow the redirect to consent.
Claude Desktop / Claude.ai: Settings → Connectors → Add custom connector → URL = https://mcp.linkupapi.com/mcp. Click Connect, approve on the LinkupAPI consent page.
If you don't have a LinkupAPI account yet, the consent page will let you sign up before approving.
Each skill writes to a shared local directory inside the project where Claude Code runs:
./campaigns/{date}-{slug}.md— outreach + high-intent + feed-engage campaign logs./enrichments/{date}-{slug}.md— enrichment logs (and the CSV/JSON outputs alongside)
Subsequent runs of any skill dedupe against these logs, so a profile is never invited, commented on, or enriched twice unnecessarily. The logs are the source of truth for accept-rate analysis 7-14 days after a campaign.
LinkupAPI is built around an action-based, multi-channel interface — the same {action, params} shape that powers linkedin_profiles / linkedin_network / linkedin_content will extend to other B2B channels as they land. The skills in this repo are LinkedIn-shaped today; the LinkupAPI surface itself is broader.
Private use only (for now).