KOLens MCP server — influencer data for Claude & AI agents

Why MCP instead of the REST API

MCP is the open protocol AI assistants use to call external tools. Connecting KOLens over MCP means Claude can run the whole research loop itself — search a niche, vet the results, build a shortlist, draft outreach — from one natural-language instruction, choosing the right tools and chaining them without you writing any integration code. The REST API (see the API reference) remains the right choice for custom software; MCP is the right choice for agents and assistants.

Connect from Claude.ai (no install)

1. In Claude.ai, open Settings → Connectors → Add custom connector.
2. Paste the connector URL: https://kolens.ai/mcp
3. Claude opens the KOLens consent page — approve with your KOLens account. The OAuth 2.1 flow (dynamic client registration + PKCE) issues a scoped token automatically; you never copy a secret.
4. Ask Claude something like: “Find 50 TikTok creators for ‘home fitness’, keep the ones with email and ≥4% engagement, and add the top 10 to a list called Q3 Fitness.”

Connect from Claude Code

claude mcp add --transport http kolens https://kolens.ai/mcp

Claude Code triggers the same OAuth consent in your browser on first use. After that, the KOLens tools are available in every session — useful for research scripts, scheduled agent runs, or wiring KOL data into a build pipeline.

Connect any other MCP client (API key)

Self-configured clients (Cursor, custom agents, MCP SDKs) can skip OAuth and send a Bearer token instead. Create a key in the web app under Account → API keys, then point your client at the endpoint with an Authorization header:

{
  "mcpServers": {
    "kolens": {
      "type": "http",
      "url": "https://kolens.ai/mcp",
      "headers": { "Authorization": "Bearer kol_xxxxxxxxxxxx" }
    }
  }
}

The 26 tools

Tool names are stable identifiers — safe to reference in prompts, skills and agent code. Grouped by workflow:

OAuth tokens carry scopes: read:kols (all read tools), write:lists (watchlist, lists, alerts, CRM mutations) and run:scrape (anything that spends credits). The consent screen shows which scopes a client requests.

Recipes — what to ask for

• Niche sweep: “Scrape 100 TikTok videos for ‘pet grooming’, then list the creators with email, ≥4% engagement and 10k–250k followers, sorted by average views.” (scrape_tiktok_by_keyword → get_job_status → list_kols)
• Vet before paying: “Pull @handle's profile, audience snapshot and activity summary — is the audience really US-based and is the account authentic?” (get_kol_profile → get_audience_snapshot → get_kol_activity_summary)
• Monday digest: “Summarise my watchlist changes this week and flag anyone rising fast.” (list_watchlist → list_alerts)
• Campaign check: “Compare the three tracked videos from our March campaign against the competitor's placement.” (compare_videos)
• Outreach: “Draft a collab email to @handle pitching our vegan protein brand, friendly tone.” (draft_outreach — the draft is returned to you; KOLens never sends email on your behalf.)