Datadive MCP Server

@datadive-tools/mcp

An MCP server that lets Claude (or any MCP-compatible client) query your DataDive niches, keywords, competitors, and Rank Radar data using your existing API key.

Runs locally on your machine over stdio. Your API key never leaves your machine except as the x-api-key header on requests to api.datadive.tools.

What you can ask

• "List my DataDive niches in marketplace com."
• "What's the master keyword list for niche z515cGOFg3?"
• "Who are the top competitors in niche X and what are their sales?"
• "What's my ranking juice for niche X — where can I improve?"
• "Show me my rank radars."
• "Plot the organic ranking trend for rank radar Y from 2024-03-01 to 2024-04-01."
• "Run a niche dive on ASIN B08N5WRWNW in the US marketplace with 5 competitors."
• "Refresh niche X with today's data, same competitors."
• "Re-dive niche X with 12 competitors but keep B08N5WRWNW in the set."
• "Is my dive done yet?"
• "Start a rank radar tracking 10 keywords for ASIN B08N5WRWNW in niche X."
• "Which Amazon seller accounts are connected?"
• "Search my catalog for active 'widget' products in the US."
• "What price or content changes happened on my listings last week?"

1. Get a DataDive API key

1. Sign in at https://2.datadive.tools.
2. Go to Settings → API Key (/api-key).
3. Click Generate API Key. Copy the value — you'll paste it into your MCP client config below.

Requires the Billing Manager or Owner role and the Standard plan or higher. Contact your org admin if you can't see the page.

2. Add it to your MCP client

Claude Desktop

Edit your config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add the datadive entry under mcpServers:

{
  "mcpServers": {
    "datadive": {
      "command": "npx",
      "args": ["-y", "@datadive-tools/mcp"],
      "env": {
        "DATADIVE_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Restart Claude Desktop. You should see datadive in the tools menu.

Claude Code

claude mcp add datadive -- npx -y @datadive-tools/mcp
# Then add the env var via:
#   claude mcp add datadive --env DATADIVE_API_KEY=YOUR_API_KEY -- npx -y @datadive-tools/mcp

Or edit .mcp.json in your project / ~/.claude/mcp.json globally with the same JSON shape as above.

Cursor

Settings → MCP → Add new MCP server and paste the same JSON shape:

{
  "datadive": {
    "command": "npx",
    "args": ["-y", "@datadive-tools/mcp"],
    "env": { "DATADIVE_API_KEY": "YOUR_API_KEY" }
  }
}

3. Verify

Ask Claude: "List my DataDive niches."

You should see a tool call to list_niches and a JSON response with your niches, plus pagination metadata. If you don't, see Troubleshooting below.

Available tools

All data is scoped to the organization that owns the API key. Most tools are read-only. The tools that change something are marked above; the ones that cannot be undone also require an explicit confirm: true — see Tools that change something.

Tools that change something

Six tools cannot be undone, so they require an explicit confirm: true argument. Called without it, they change nothing and return a note asking the assistant to confirm the cost with you first:

• create_niche_dive, redive_niche — spend dive tokens (scales with numberOfCompetitors).
• create_rank_radar — spends Search Term tokens (scales with numberOfKeywords).
• generate_listing_copy — spends one AI Copywriter prompt per call.
• delete_niche, delete_rank_radar — destroy data permanently.

Check your remaining balance any time with get_quota.

The other write tools — pause_rank_radar, resume_rank_radar and the three *_rank_radar_search_terms tools — need no confirmation. They only move Daily Tracked Keywords capacity, which is freed again when you pause, and each is undone by its counterpart.

To skip the per-call confirmation (e.g. in an automated setup), set DATADIVE_AUTO_CONFIRM_WRITES=true in your client config — then all six run without confirm, deletes included.

None of the three is safe to retry — each call spends tokens again and creates a separate dive / re-dive / Rank Radar, even with identical arguments. If a call errors or times out, check get_dive_status / list_niches / list_rank_radars for what already exists before calling again.

Dives are asynchronous. create_niche_dive returns a diveId and an estimated completion time immediately; poll get_dive_status with that diveId until it reports success, which carries the new nicheId you then feed to list_niches, get_niche_keywords, and the other niche tools.

redive_niche refreshes a niche you already have rather than creating another one, and it keeps the same nicheId — so rank radars and reports built on that niche follow the refreshed data. It takes a mode:

• same_competitors — re-dive the niche's current competitor set. Nothing else to supply; use it purely to pull in current sales, price and keyword data.
• discover — search for a fresh competitor set of numberOfCompetitors ASINs. Steer it with heroAsin (the seed to build around; defaults to the niche's highest-selling competitor, preferring one of your own ASINs), lockedAsins (kept no matter what discovery finds) and excludedAsins (never selected).

⚠️ For create_niche_dive, marketplace uses full Amazon domain suffixes (com, co.uk, com.mx, co.jp, …) — e.g. the UK marketplace is co.uk, not uk.

Configuration

Troubleshooting

If a tool call returns an error message, it'll be one of these — each maps to a specific HTTP status from the DataDive API.

If the server itself fails to start, look in your MCP client's log output:

• DATADIVE_API_KEY environment variable is required — the env var isn't being passed through to the binary. Confirm your client config has it under env (not args), and that you've restarted the client after editing.

Privacy

• The MCP server runs locally on your machine. It is a thin shim — every tool call becomes a single HTTPS request from your machine to api.datadive.tools with your API key as the x-api-key header.
• Your API key is stored only in your MCP client's config file (which you control). It is never sent anywhere else.
• The server logs nothing on its own. Standard backend access logs at DataDive record per-request metadata (organization, route, status, latency) the same way any direct API call would.

Development

npm install
npm run build       # tsup -> dist/index.js (with shebang)
npm test            # vitest run
npm run typecheck   # tsc --noEmit
npm run lint        # eslint

Test the binary end-to-end with the MCP Inspector:

DATADIVE_API_KEY=YOUR_API_KEY npx @modelcontextprotocol/inspector dist/index.js

Releasing

The repo uses Changesets:

npx changeset             # add a changeset describing the change (patch/minor/major)
git commit -am "fix: ..."
git push

A "Version Packages" PR opens automatically. Merging it bumps the version, updates CHANGELOG.md, and publishes to npm with provenance.

Support

• Issues: https://github.com/Data-Dive-Tools/datadive-mcp/issues
• API docs: https://developer.datadive.tools/docs#/v1
• DataDive support: support@datadive.tools

License

MIT — see LICENSE.