rows.page MCP server

rows.page is a place for your rows. Your agent pushes CSV, TSV, JSON, JSONL or Parquet with one tool call and gets back a link a human can explore: filter, chart columns, run SQL, read pinned findings, and see what changed since the last push. Viewing is free, no signup needed. Anonymous datasets live 7 days unless someone keeps them.

The server is remote only: https://rows.page/mcp, streamable HTTP, stateless JSON-RPC. There is nothing to install. A GET on the endpoint returns the server card.

Connect

Claude Code

claude mcp add --transport http rows https://rows.page/mcp

Claude Desktop (claude_desktop_config.json, needs Node.js)

{
  "mcpServers": {
    "rows": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://rows.page/mcp"]
    }
  }
}

Cursor (.cursor/mcp.json)

{
  "mcpServers": {
    "rows": {
      "url": "https://rows.page/mcp"
    }
  }
}

VS Code (.vscode/mcp.json)

{
  "servers": {
    "rows": {
      "type": "http",
      "url": "https://rows.page/mcp"
    }
  }
}

Windsurf (mcp_config.json)

{
  "mcpServers": {
    "rows": {
      "serverUrl": "https://rows.page/mcp"
    }
  }
}

Codex CLI (~/.codex/config.toml)

[mcp_servers.rows]
url = "https://rows.page/mcp"

Gemini CLI (~/.gemini/settings.json)

{
  "mcpServers": {
    "rows": {
      "url": "https://rows.page/mcp",
      "type": "http"
    }
  }
}

API key (optional)

Anonymous pushes work. An API key (created at https://rows.page/account) pushes into your account, unlocks list_datasets, and lets you re-push by a stable key. Send it as Authorization: Bearer rpk_... or append ?apiKey=rpk_... to the URL.

claude mcp add --transport http rows https://rows.page/mcp --header "Authorization: Bearer rpk_..."

Tools

Tool	What it does
`push`	Push `csv`, `tsv`, `json`, `jsonl`, `rows` (array of objects) or a `url`. Optional `title`, `summary`, `key`, `id_column`, `findings[]` and provenance (`source`, `run`, `commit`). Pass `dataset` + `token` for a new version. Returns the link plus next steps for your human.
`pin_finding`	Pin a saved view (title, note, `q` filter, `sort`, `cols` or SQL) to a dataset. The page opens on its findings.
`get_dataset`	Meta, versions, findings, columns and the first 20 rows.
`list_datasets`	Your datasets, newest first. Needs an API key.
`delete_dataset`	Delete a dataset you own. Its URL stops working immediately.
`get_limits`	Your plan, its caps and usage.

Every push is an immutable version. Identical bytes return unchanged: true. Re-pushes report rows new, gone and changed (set id_column so changed rows can be matched) plus a diff URL. Finding filters use the rows.page query language (status:failed latency_ms:>2000, (country:NL OR country:BE) amount:>500) or SQL starting with SELECT.

Example prompts

Push results.csv to rows.page titled "Eval run 412", pin the failed rows as a finding with a one-line note on what they have in common, and give me the link.

Query the last 24h of orders, push them to rows.page with key nightly-orders and id_column order_id, then tell me how many rows changed since yesterday.

Fetch https://example.com/export.parquet, push it to rows.page, pin a finding for amount:>500 sorted by -amount, and hand me the owner link.

Limits

Anonymous: 100 MB per file (16 MB inline per MCP push, use url for bigger files), deleted 7 days after creation unless kept, 50 pushes and 1 GB per day per IP, 60 MCP calls per minute. Free, Pro and Max plans: https://rows.page/pricing.

rows.page MCP server