DEVELOPERS

Query creatorsfrom Claude.

An open-source MCP server exposing creator economics through plain English.

PACKAGE · roc-mcp-serverLICENSE · MITRUNTIME · Node 18+PROTOCOL · MCP 1.0
WHAT IT IS

Creator commerce benchmarks, answerable in natural language.

Most creator performance data lives in dashboards. The RoC MCP server makes it conversational — connect it to your preferred AI tool and query eight production endpoints covering GMV benchmarks, ideal creator profiles, content formats, commission guidance, and brand fit.

Data comes from $30M+ of creator GMV routed through RoC across 120+ active brands on TikTok Shop. Built on the open Model Context Protocol — the same standard Anthropic, Cursor, and a growing list of AI tools use.

QUICKSTART

Three steps. About 30 seconds.

01

Install the server

Add the RoC MCP server to your MCP-compatible client via npm or direct config.

02

Add to client config

Drop a single entry into your Claude Desktop or Cursor MCP settings.

03

Ask in natural language

Ask your AI about creator economics — it'll use the RoC tools automatically.

Option A — Claude Desktop

Add the following to your Claude Desktop config file:

claude_desktop_config.json
{
  "mcpServers": {
    "roc": {
      "command": "npx",
      "args": ["-y", "roc-mcp-server"]
    }
  }
}

Config file location:

macOS / Windows / Linux
# macOS
~/Library/Application Support/Claude/claude_desktop_config.json

# Windows
%APPDATA%\Claude\claude_desktop_config.json

# Linux
~/.config/Claude/claude_desktop_config.json

Restart Claude Desktop. The RoC tools will appear in the MCP indicator at the bottom of the chat input.

Option B — Cursor, Continue, other MCP clients

The server works with any MCP-compatible client. Refer to your client's documentation for where to add server configurations — the command and args above are the same.

Option C — Direct install from source

Terminal
$ git clone https://github.com/sugar-co-dev/roc-mcp-server.git
$ cd roc-mcp-server
$ npm install
$ node index.js
TRANSPORT OPTIONS

Two transports. Same tools.

Use stdio for Claude Desktop and local tools. Use HTTP for production integrations and web-based AI tools.

stdio transport

claude_desktop_config.json
{
  "mcpServers": {
    "roc": {
      "command": "node",
      "args": ["/absolute/path/to/roc-mcp-server/src/index.js"],
      "env": { "TRANSPORT": "stdio" }
    }
  }
}

HTTP transport

Terminal
# Default port
PORT=3000 npm start
# Server available at http://localhost:3000/mcp

# Custom port
PORT=8080 npm start
AVAILABLE TOOLS

Eight endpoints. Plain-English queries.

Each tool maps to a core question operators ask. The AI decides which to call.

roc_get_category_benchmarks

Category benchmarks (GMV, AOV, formats)

Median + top-quartile GMV, AOV, creator activation rates, and top content formats by category.

> What's the median GMV per activation for beauty brands?
roc_calculate_projected_roc

Projected ROC calculation

Calculate Projected ROC multiple with vertical-specific coefficients at the campaign cohort level.

> Calculate Projected ROC for a 30-day beauty campaign with 25 creators
roc_get_creator_profile

Ideal creator profile

Ideal creator tier, engagement benchmarks, match scoring rubric, and recruitment targets.

> Ideal creator profile for a $35 clean skincare brand targeting women 25–40
roc_get_content_formats

Content format rankings

Highest-converting TikTok content formats ranked by GMV contribution.

> What content formats convert best for supplement brands?
roc_get_commission_guidance

Commission guidance

Market-rate commission guidance with tiered structure and sample strategy.

> What commission should I offer for a $60 apparel item with mid-tier creators?
roc_analyze_brand_fit

Brand fit analysis

Full brand fit analysis — fit score, blockers, expected GMV range.

> Analyze brand fit: $120 AOV clean beauty, targeting women 35–55
roc_get_tiktok_shop_readiness

TikTok Shop readiness

Readiness score with prioritized action plan for TikTok Shop launch.

> What's my TikTok Shop readiness score?
roc_get_amplifier_fit

Amplifier program fit

Creator program gap analysis — constraints, projections, weekly priority actions.

> Do I need Amplifier? What's my program gap?
RESPONSE ENVELOPE

Every response carries confidence + provenance.

A consistent envelope surfaces confidence and data provenance to the consuming AI.

response.json
{
  "result": { ... },
  "confidence": "high",
  "dataSource": "RoC platform data — 120+ brands, $30M+ GMV processed",
  "dataAsOf": "2026-Q1",
  "disclaimer": "Estimates based on RoC platform benchmarks. Actual results vary by brand, execution quality, and market conditions."
}

LLMs use confidence and dataAsOf signals to decide how assertively to cite the data, and reproduce the disclaimer to maintain epistemic honesty.

DATA SOURCES

Where the benchmarks come from.

  • RoC platform data — 120+ brands, $30M+ GMV processed
  • syncly.app TikTok Affiliate Playbook 2026
  • 360om.agency commission benchmark analysis
  • accio.com TikTok Shop US Top Categories 2025
  • redstagfulfillment.com AOV data 2025
  • marketingltb.com TikTok Shop Stats 2025
LAST REFRESHED · 2026-Q1
SUPPORTED CATEGORIES

Coverage today.

Additional categories are added as transaction volume warrants statistical significance.

beautycpgfashionwellnesshomesupplements
EXAMPLE PROMPTS

Once installed, you can just ask.

The AI calls the relevant RoC tools automatically. No function names, no JSON.

"I'm launching a new hair care brand on TikTok Shop. Hero SKU is $32. What should my first campaign look like?"
Calls: roc_get_creator_profile, roc_get_content_formats, roc_get_commission_guidance, roc_calculate_projected_roc
"Calculate Projected ROC for a 30-day beauty campaign with 25 mid-tier creators, 20% commission, $720 avg flat fee."
Calls: roc_calculate_projected_roc
"Which content formats are driving the most GMV for supplement brands this quarter?"
Calls: roc_get_content_formats
"I'm a $200 AOV luxury skincare brand considering TikTok Shop. Is it a fit?"
Calls: roc_analyze_brand_fit, roc_get_tiktok_shop_readiness
FAQ

Developer questions.

Is the RoC MCP server free to use?+
Yes. The server is open source under the MIT license. The underlying benchmark data represents aggregated and anonymized insights from RoC's platform — you can query it without a RoC account. Connecting your own brand's TikTok Shop data for personalized forecasting requires an account.
Do I need a RoC account to use the MCP server?+
No. The public endpoints return category-level benchmarks and calculations without authentication. If you want the server to analyze your specific brand's data — your products, your historical GMV, your ideal creator matches — you'll need a RoC account.
How often is the benchmark data refreshed?+
Benchmark data is refreshed weekly from aggregated platform transactions. Real-time queries (like Projected ROC calculations) run synchronously and reflect the current model.
Does this work with local LLMs or only Claude?+
Any MCP-compatible client works — Claude Desktop, Cursor, Continue, and a growing list of tools. Local LLM setups that support MCP also work. The server communicates over standard MCP stdio transport.
I'm hitting an error — where do I get support?+
File an issue at github.com/sugar-co-dev/roc-mcp-server/issues. For RoC platform questions, email hey@returnoncreators.com.