Build creator marketing agents on real data

Build creator marketing agents on top of Creatorland's creator and market intelligence data set. Use one MCP connection to ground pricing, casting, outreach, and analytics workflows in real creator and deal data.

Creatorland MCP handles the data layer. Your agent handles the workflow.

The runnable source lives in examples/grounded-agent. It connects to Creatorland MCP, pulls a deal-derived rate band, and can ask Claude to answer using only that data instead of guessing.

Grounded rate-band example

import { Client } from '@modelcontextprotocol/sdk/client/index.js'
import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js'

const transport = new StreamableHTTPClientTransport(new URL('https://mcp.creatorland.com/mcp'), {
  requestInit: { headers: { Authorization: `Bearer ${process.env.MCP_TOKEN}` } },
})
const client = new Client({ name: 'grounded-agent-example', version: '0.1.0' })
await client.connect(transport)

// Real, deal-derived economics for a creator slice. Privacy-safe by construction.
const res = await client.callTool({
  name: 'query_market_intelligence',
  arguments: { mode: 'rate', vertical: 'Beauty', creator_tier: 'micro' },
})
console.log(JSON.parse(res.content[0].text))

That is the grounding step. Hand the result to any LLM with instructions to answer using only the returned data, and your agent can quote defensible numbers instead of guessing.

Run it

git clone https://github.com/creatorland/creatorland-data-mcp-site
cd creatorland-data-mcp-site/examples/grounded-agent
npm install
MCP_TOKEN=<your token> node index.mjs "what should a beauty micro-influencer charge?"

Add ANTHROPIC_API_KEY=<key> to see Claude answer using the returned rate band.

Getting a token

Creatorland MCP signs you in through a free Creatorland account. Connect it to any supported client, sign in, and use the session token for local examples.

Most clients run OAuth for you. This example accepts a bearer token directly to keep the demo small. In production, use your client or SDK's OAuth support instead of handling raw tokens manually.

Another example: Competitor Roster X-Ray

Enter a brand and get two lists back: creators with confirmed affiliations, and lookalike whitespace the brand has not locked up yet.

Confirmed affiliations can come from real deals, declared partnerships, or detected brand activity. This is a real affiliation graph, not hashtag-scraping inference.

The runnable source lives in examples/competitor-roster-xray. It composes two calls to one tool:

// 1) The brand's known roster: confirmed affiliations only.
const roster = await client.callTool({
  name: 'search_creators',
  arguments: { mode: 'brief', filters: { brand_affiliations: ['Fenty Beauty'] }, limit: 25 },
})

// 2) Whitespace: lookalikes of the strongest affiliated creator, minus the roster.
const similar = await client.callTool({
  name: 'search_creators',
  arguments: {
    mode: 'lookalike',
    seed_creator: { type: 'creatorland_user_id', id: seedId },
    limit: 25,
  },
})

Whitespace means similar creators minus the known roster. Treat it as a casting shortlist, not a guarantee that a creator is unaffiliated. Confirm any individual creator with get_creator_profile before you pitch.

Another example: Creator Pricing Mirror

Most creators price without good market context. This example uses Creatorland's creator and market intelligence data set to compare a creator's current rate against real deals for similar creators in the same vertical and follower tier.

The runnable source lives in examples/creator-pricing-mirror. It derives the creator's tier from their profile, then positions their rate in a real band:

const m = await client.callTool({
  name: 'query_market_intelligence',
  arguments: { mode: 'rate', vertical: 'Beauty', creator_tier: 'micro', quoted_rate: 120 },
})
// -> { p25_usd, median_usd, p75_usd, quoted_rate_position, provenance, confidence }

The band is computed across creators in the tier, not from any one creator. It comes from real deal data, not a survey estimate.

The read tools you can build with

These examples call one or two tools, but the same pattern works across the Creatorland MCP read tools:

• search_creators
• get_creator_profile
• query_market_intelligence
• get_member_content_stats

get_member_content_stats is entitlement-gated and only works for consented members you work with directly.

See the tools reference for inputs, response shapes, and credit costs.

To turn a search_creators result into a rendered creator card, follow the build a creator card recipe.