arpeio-mcp

--- title: Arpeio MCP emoji: 🛠️ colorFrom: blue colorTo: indigo sdk: docker pinned: false ---

arpeio-mcp

<!-- mcp-name: io.github.arpe-io/arpeio-mcp -->

 

Unified MCP server for Arpe.io data tools — build, preview, and execute high-performance data commands through AI assistants.

| Tool | Description | |------|-------------| | FastBCP | High-performance parallel database export to files and cloud | | FastTransfer | High-performance parallel data transfer between databases | | LakeXpress | Automated database-to-cloud data pipeline as Parquet | | MigratorXpress | Cross-platform database migration with parallel transfer |

No binaries required. All tools work in command builder mode out of the box — command building, preview, and informational tools work without any Arpe.io binary installed. To enable execution, download the binaries from arpe.io and set the corresponding *_PATH environment variables.

Connect your AI assistant

A hosted instance is available at https://arpe-io-arpeio-mcp.hf.space/sse — no installation required. For local installation with execution support, use the stdio transport via pip install arpeio-mcp.

ChatGPT | Claude Code | Claude Desktop | Cursor | Gemini CLI | HuggingChat | Kiro IDE | Le Chat (Mistral) | VS Code | Windsurf

ChatGPT

Available for paid plans only (Plus, Pro, Team, and Enterprise).

1. Open ChatGPT in your browser, go to Settings > Apps and connectors.
2. Open Advanced settings and enable Developer mode.
3. Go to Connectors > Browse connectors > Add a new connector.
4. Set the URL to https://arpe-io-arpeio-mcp.hf.space/sse and save.

Claude Code

claude mcp add --transport sse arpeio https://arpe-io-arpeio-mcp.hf.space/sse

Or for local installation with execution support:

pip install arpeio-mcp
claude mcp add arpeio arpeio-mcp

Claude Desktop

Add the following to your Claude Desktop configuration file:

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

Remote (no installation)

{
  "mcpServers": {
    "arpeio": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://arpe-io-arpeio-mcp.hf.space/sse"
      ]
    }
  }
}

Local (with execution support)

{
  "mcpServers": {
    "arpeio": {
      "command": "arpeio-mcp",
      "env": {
        "FASTBCP_PATH": "/path/to/FastBCP",
        "FASTTRANSFER_PATH": "/path/to/FastTransfer",
        "LAKEXPRESS_PATH": "/path/to/LakeXpress",
        "MIGRATORXPRESS_PATH": "/path/to/MigratorXpress"
      }
    }
  }
}

Cursor

1. Open Cursor Settings and search for "MCP".
2. Add a new MCP server with the following configuration:

{
  "mcpServers": {
    "arpeio": {
      "url": "https://arpe-io-arpeio-mcp.hf.space/sse"
    }
  }
}

Gemini CLI

Add the following to your ~/.gemini/settings.json file:

{
  "mcpServers": {
    "arpeio": {
      "uri": "https://arpe-io-arpeio-mcp.hf.space/sse"
    }
  }
}

HuggingChat

1. In the chat interface, click the + icon, select MCP Servers, then Manage MCP Servers.
2. Click Add Server.
3. Set the Server Name to Arpe.io and the Server URL to https://arpe-io-arpeio-mcp.hf.space/sse.
4. Click Add Server and verify the health check shows Connected.

Kiro IDE

Add the following to your Kiro MCP configuration file (.kiro/settings/mcp.json in your workspace):

{
  "mcpServers": {
    "arpeio": {
      "url": "https://arpe-io-arpeio-mcp.hf.space/sse"
    }
  }
}

Le Chat (Mistral)

Available on all plans, including free.

1. Go to Intelligence > Connectors.
2. Click Add connector > Custom MCP Connector.
3. Set the name to Arpe.io and the URL to https://arpe-io-arpeio-mcp.hf.space/sse.
4. Leave authentication disabled and click Create.

VS Code (GitHub Copilot)

Add the following to your VS Code MCP configuration. Run MCP: Open User Configuration from the Command Palette to open it.

• Linux: ~/.config/Code/User/mcp.json
• macOS: ~/Library/Application Support/Code/User/mcp.json
• Windows: %APPDATA%\Code\User\mcp.json

{
  "servers": {
    "arpeio": {
      "url": "https://arpe-io-arpeio-mcp.hf.space/sse",
      "type": "sse"
    }
  }
}

Windsurf

Add the following to your Windsurf configuration file:

• Linux: ~/.codeium/windsurf/mcp_config.json
• macOS: ~/.codeium/windsurf/mcp_config.json
• Windows: %USERPROFILE%\.codeium\windsurf\mcp_config.json

{
  "mcpServers": {
    "arpeio": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://arpe-io-arpeio-mcp.hf.space/sse"
      ]
    }
  }
}

Local Installation

For full execution support (not just command building), install locally:

pip install arpeio-mcp

Then configure your AI assistant to use the arpeio-mcp command (stdio transport) with optional binary paths — see the Claude Desktop local configuration for an example.

Environment Variables

| Variable | Description | Required | |----------|-------------|----------| | FASTBCP_PATH | Path to FastBCP binary | No | | FASTTRANSFER_PATH | Path to FastTransfer binary | No | | LAKEXPRESS_PATH | Path to LakeXpress binary | No | | MIGRATORXPRESS_PATH | Path to MigratorXpress binary | No | | FASTBCP_DIR_PATH | FastBCP directory for LakeXpress | No | | FASTTRANSFER_DIR_PATH | FastTransfer directory for MigratorXpress | No | | _TIMEOUT | Per-tool execution timeout (seconds) | No | | _LOG_DIR | Per-tool log directory | No | | LOG_LEVEL | Logging level (DEBUG/INFO/WARNING/ERROR) | No |

Available Tools (17)

Read-only advisory tools were consolidated into one *_info tool per product (with an action enum) and per-product release-notes tools were merged into a single arpe_release_notes. Auto-parallelism is now suggested inside preview so the typical workflow is 2 calls (preview → execute) instead of 4.

FastBCP (3 tools)

• fastbcp_info — Read-only advisory: action="formats" | "parallelism" | "workflow" | "version"
• fastbcp_preview_export — Validate parameters and render the command (auto-suggests parallelism when method is omitted)
• fastbcp_execute_export — Run the export

FastTransfer (3 tools)

• fasttransfer_info — Read-only advisory: action="combinations" | "parallelism" | "workflow" | "version"
• fasttransfer_preview_transfer — Validate parameters and render the command (auto-suggests parallelism)
• fasttransfer_execute_transfer — Run the transfer

LakeXpress (3 tools)

• lakexpress_info — Read-only advisory: action="capabilities" | "workflow" | "version"
• lakexpress_preview_command — Build any LakeXpress command (lxdb_, config_, sync, sync[export], sync[publish], run, status, cleanup). On v0.4.0+ binaries, warns when -a / --lxdb_auth_id / --sync_id are missing on sync-family calls
• lakexpress_execute_command — Run the command

MigratorXpress (4 tools)

• migratorxpress_info — Read-only advisory: action="capabilities" | "workflow" | "version"
• migratorxpress_validate_auth_file — Validate the JSON auth file (only file-I/O advisory tool kept separate)
• migratorxpress_preview_command — Build the migrate command. Accepts the new project tag (v0.6.30+); warns on migration_db_type="postgres" against pre-0.6.32 binaries
• migratorxpress_execute_command — Run the migration

Meta (4 tools)

• arpe_get_status — Status of all four CLIs (installed / command-builder-only)
• arpe_quick_start — Detect the right tool from a plain-English use case and return a workflow guide
• arpe_release_notes — Return release-notes chunks for any product (product="fastbcp" | "fasttransfer" | "lakexpress" | "migratorxpress", optional version)
• search_docs — BM25 full-text search over arpe.io docs sites and blog

Structured output

The command-building, execution, and discovery tools return structured content (outputSchema + structuredContent) alongside the human-readable markdown, so MCP clients can chain calls programmatically instead of parsing prose:

• *_preview_ → the built command (argv), command_string, masked command_display, explanation, and version warnings. The exact command can be handed straight to the matching _execute_* tool.
• *_execute_* → success, return_code, stdout, stderr, and parsed diagnostics.
• search_docs → ranked results records; arpe_get_status → per-product status; arpe_release_notes → release-notes chunks.

Each payload carries a status field (ok / error) so success and error responses are both machine-checkable.

Prompts (5)

Conversation starters surfaced by clients that support MCP prompts (Claude Desktop, Cursor, etc.): export-table, transfer-data, lakehouse-pipeline, migrate-database, troubleshoot.

Resources (4)

Static capability matrices served as MCP resources so clients can prefetch them without a tool call: arpeio://capabilities/fastbcp-formats, fasttransfer-combinations, lakexpress-capabilities, migratorxpress-capabilities.

Evaluations

evaluations/arpeio_eval.xml holds 10 read-only, verifiable questions used to check that an LLM can drive the server to correct answers (tool selection, parallelism recommendations, capability lookups, version gating). A guard test re-derives every answer from the capability registries so the answer key stays honest:

python -m pytest tests/test_evaluations.py -q

License

MIT