Claude skill, subagent & MCP
Let your AI assistant run your LinkedIn, with the same approval gate.
Claude skill
One skill that bundles every playbook (hooks, writing, humanizer, profile, career, engagement, networking, company page, analytics, publishing) and routes each request to the right one.
- Claude apps: download linkedin-agent-skill.zip, then Settings → Capabilities → Skills → Upload.
- Claude Code:
cp -r skill/linkedin-agent ~/.claude/skills/
Subagent (Claude Code)
cp agents/linkedin-agent.md ~/.claude/agents/Then ask: “Use the linkedin-agent subagent to plan and draft my week.”
MCP server
{"mcpServers": {"linkedin-agent": {"command": "linkedin-agent", "args": ["mcp"]}}}
Claude Code: claude mcp add linkedin-agent -- linkedin-agent mcp. Over HTTP: linkedin-agent mcp --http.
25 tools
status, list_skills, run_skill, get_skill_prompt, save_draft, list_posts, approve_post, publish_post, attach_media, render_carousel_pdf, generate_post_image, record_metrics, get_analytics, refresh_trends, get_brand_profile, update_brand_profile, add_story, list_documents, save_document, list_persona_packs, apply_persona_pack, import_linkedin_export, import_analytics_file, list_workspaces, switch_workspace
publish_post refuses anything that hasn't been approved, and the skill tells the model to approve only after you say yes to the exact text.