How to add flyto-core to Claude Desktop

A debuggable automation engine for AI agents — trace every step, replay from any point, 467 modules as MCP tools. Paste the config into ~/Library/Application Support/Claude/claude_desktop_config.json and restart Claude Desktop.

Last updated June 14, 2026 · 274★ · stdio · no auth

Claude Desktop config for flyto-core

pip install flyto-core

{
  "mcpServers": {
    "flyto-core": {
      "command": "python",
      "args": [
        "-m",
        "core.mcp_server"
      ]
    }
  }
}

Setup steps

1Open Claude Desktop → Settings → Developer → Edit Config (this opens ~/Library/Application Support/Claude/claude_desktop_config.json).
2Paste the flyto-core config below under the top-level "mcpServers" key.
3Fill in any placeholder secrets (API keys, paths) in the snippet.
4Save the file, then fully quit and reopen Claude Desktop.
5Open a chat and confirm flyto-core's tools appear under the 🔌 tools menu.

Before you start

Python 3.9+
pip install flyto-core (core engine + CLI + MCP server)
For browser modules: pip install flyto-core[browser] then playwright install chromium
API keys / credentials (via environment variables) for integration modules such as OpenAI, Anthropic, Gemini, Notion, Slack, AWS S3, Google

What flyto-core can do in Claude Desktop

search_modules

Discover modules using word-level and tag matching scoring for relevant natural-language queries across the 467-module catalog.

validate_params

Validate and auto-correct module parameters; maps wrong field-name aliases (e.g. 'site' to 'url') and suggests alternatives when a non-existent module is requested.

browser.*

38 browser-automation modules: launch (with channel: 'chrome' support), goto, click, fill forms, extract (text mode by default), evaluate, screenshot, viewport, performance/Web Vitals, wait, close.

flow.*

24 flow-control modules: switch, loop, branch, parallel, retry, circuit breaker, rate limit.

api.*

11 API integration modules: OpenAI, Anthropic, Gemini, Notion, Slack, Telegram.

file.*

8 file modules: read, write, copy, move, delete, exists, edit, diff.

data.*

8 data modules: parse and generate JSON, XML, YAML, and CSV.

image.*

9 image modules: resize, convert, crop, rotate, watermark, OCR, compress.

Security

Apache-2.0 licensed. The MCP server exposes powerful modules including sandbox code execution (Python/Shell/JavaScript), shell/SSH access, Docker and Kubernetes control, file system read/write, and network scanning — grant access deliberately. Many integration modules (OpenAI, Anthropic, AWS S3, Google, Slack, etc.) require their own API keys/credentials supplied via environment variables. Report vulnerabilities to security@flyto.dev.

flyto-core + Claude Desktop FAQ

Where is the Claude Desktop config file?

Claude Desktop reads MCP servers from ~/Library/Application Support/Claude/claude_desktop_config.json. Paste the flyto-core config there under the "mcpServers" key and restart the client.

Is flyto-core safe to use with Claude Desktop?

What does the MCP server actually expose?

All of flyto-core's modules become MCP tools — the README states 467 modules across 78 categories, so 'your AI gets all modules as tools', plus helper tools like search_modules and validate_params.

How do I add it to Claude Code?

Install with `pip install flyto-core`, then run `claude mcp add flyto-core -- python -m core.mcp_server`, or add the mcpServers JSON block with command 'python' and args ['-m', 'core.mcp_server'] to your MCP config.

Is there an HTTP/remote transport?

Yes. `pip install flyto-core[api]` then `flyto serve` exposes an HTTP API including a POST /mcp endpoint using the MCP Streamable HTTP transport. A hosted deployment is also available on Fronteir AI.

View repo Full flyto-core page